Note: Descriptions are shown in the official language in which they were submitted.
INS DECODER
FIELD OF ~ INVENTION
This application is a division of Canadian Patent Application
No. 2,303,514 filed on September 3, 1998.
The present invention generally relates to the field of digital
image processing and, more specifically, to an image decoding
device for decoding coded data with high efficiency encoded by an
image coding device.
BACKGROUND OF THE INVENTION
Flash Pix forniat specification version 1.0 has been proposed
as an image forniat for converting natural image data into digital
data suitable for computer processing.
.This format specification permits a plurality of data with
different resolutions to be stored together therein so that any
data suited to an actual display and/or printing device can be
selected and taken-out promptly in response to a user's request.
Furthernwre, each image is divided into tiles arranged in the
format that allows the user to select only a necessary data portion
of the image and process it in an enlarged or reduced size with a
reduced processing load.
Referring to Figures 1 and 2, an image coding device for
encoding an image according to the flash pix forn~at is described as
follows. In Fig. 1, images are shown in different reduced
1
scales, each of which are divided into tiles. Figure 2 is a
block diagram of an exemplary image coding device.
The flash pix method is featured in that it generates first
images 1 to 4 in sizes 1/1 to 1/8, as shown in Fig. 1, then
divides each image into tiles and compresses data of each tile
image.
First, a case of encoding an image 1 shown in Fig. 1 by
the coding device of Fig. 2 is described. In Fig. l, a dashed
line shows the boundary between tiles.
A tile decomposition portion 11 divides an original image
into tiles each comprising 64 x 64 pixels, and the tiles are then
compressed one by one by a JPEG compressor portion 12. In a
coded-data integration portion 13, the coded data of each tile is
combined with tile decomposition, information from the tile
decomposition portion I1 to form coded data 1 to be output.
The image 2 of Fig . 1 is described . The original image 0
is reduced to 1/2 in length and width by a 1/2 contraction
portion 14, and then the 1/2-size image is processed through
a tile decomposition portion 15, a JPEG compressing portion
16 and a coded-data integration portion 17 to form coded data
2.
Size reduction of the image to generate a group of
size-reduced images in Fig 1 (Images 2 to 4) is repeatedly
performed until a downsized image containable within a single
tile is obtained. For example, the image 3 is still larger than
a tile and is further contracted by a factor of 2 to obtain
2
the image 4 allowable within a single tile as shown in Fig.
1. The size-reduction procedure is now finished.
Coded data for the image 3 is produced through a 1/2
contraction portion 18, a tile decomposition 19, a JPEG
compressing portion 20 and a coded data integration portion
21. Coded data for the image 4 is produced through a 1/2
contraction portion 22 , a tile decomposition portion 23 , a JPEG
compressing portion 24 and a coded data integration portion
25.
However, the above-described system involves the following
problems: Storing coded data for images downsized with
different resolutions in addition to coded data for the image
with the scale 1: 1 results in increasing a volume of coded data
by a factor of 1 .4. Furthermore, compression for encoding data
must be done for each resolution image, resulting in
considerably increasing the processing load.
On the other hand, apart from the Flash Pix method
image compression can be also accomplished by the wavelet
transform technique whereby image data with different
resolutions can be easily decoded from coded and compressed
data of an original-size image. This technique is therefore
free from the problem of increasing the amount of coded data.
Namely, the wavelet transform method can meet the demand
for decoding data with different resolutions without any
increase of coded data whereas the Flash Pix method has an
increase by a factor of 1.4 in volume of coded data.
3
Figure 3 is a basic block diagram of a wavelet transform
coding portion wherein an original image is converted by a
wavelet transform portion 31 into data for subband divisions ,
which data is quantized by a quantizing portion 104 and then
entropy encoded by an entropy coding portion 33 to produce coded
data. The wavelet transform portion 31, quantizing portion 32
and entropy coding portion 33 composes a so-called wavelet
coding portion 34.
Figure 4 is a detailed block diagram of the wavelet
transform portion 31 of Fig. 3.
Figure 5 depicts an example of the wavelet transformation
of an image. Figures 4 and 5 are shown as an example of conducting
two-dimensional subband decomposition three times.
An original image shown in Fig. 5A is filtered through a
horizontal low-pass filter 4l~and a horizontal high-pass filter
42 to create two horizontal subbands that are then decimated
to 1/2 respectively by 1/2-subsampling portions 47 and 48.
Two horizontally divided subbands are divided each into
two subbands through vertical low-pass filters 43, 45 and
vertical high-pass filter 44, 46, which subbands are decimated
each to 1/2 by 1/2 sampling portions 49 to 52. Consequently,
four subbands are formed.
A high-horizontal and high-vertical frequency subband j
(Fig.4), a high-horizontal and low-vertical frequency subband
i (Fig. 4) and a low-horizontal aid high-vertical frequency
subband h (Fig. 4 ) correspond to wavelet transform coefficients
4
h, a and j (Fig. 5B) respectively.
After this, only a remaining low-horizontal and low-
vertical frequency subband 53 is recursively divided into
subbands.
This recursive subband decomposing process is performed
by horizontal low-pass filters 54, 66, horizontal high-pass
filters 55 , 67 , vertical low-pass f alters 56 , 58 , 68 , 70 ,
vertical high-pass filters 57, 59, 69, 71 and 1/2-sampling
portions 60-65, 72-77.
Sub-bands-a-g (Fig. 4) correspond to sub-bands a-g (Fig.
5B) respectively.
Wavelet transform coefficients shown in Fig. 5B are
quantized on a subband-by-subband basis by a quantizing portion
32 (Fig. 3) and then entropy encoded by an entropy coding
portion 33 to produce coded .data. The entropy-coding portion
33 may use Huffman coding or arithmetic coding.
On the other hand, wavelet-coded data is decoded by an
entropy decoding portion 81 and inversely quantized by an
inverse quantizing portion 82. Subbands are then combined by
an inverse wavelet transform portion 83 to produce a decoded
image. The entropy decoding portion 81, inverse quantizing
portion 82 and inverse wavelet transform portion 83 compose
a so-called wavelet decoding portion 84.
Image-encoding using the wavelet transform technique is
featured by hierarchical structure according to resolution
levels as shown in Fig . 5B . This method can easily decode images
having different resolution levels from part of the coded
data or a whole of the coded data.
Namely, an image of a quarter ( 1/4 ) the original image size
can be decoded by decoding subbands a, b, c and d. An image
of a half ( 1/2 ) the original image size can be decoded by decoded
subbands a, b, c, d, e, f and g. A complete (1/1) size image
can be produced by decoding all subbands.
Referring to Fig. 7, the operation of the horizontal
low-pass (H-LP), horizontal high-pass (H-HP), vertical
low-pass (V-LP) and vertical high-pass (V-HP) filters shown
in Fig. 4 will be described as follows. Figure 7B is an enlarged
view of an encircled part B' of Fig. 7A.
When an output of ~a horizontal 9 tap filter, associated
with a pixel 91 positioned right top on the original image is
calculated for wavelet transformation of an original image,
the operation of the filter must be performed on an area 92.
However, a part of the objective area 92 is out of the
boundary of the original image, where no data exists. The
vertical filters may also encounter a similar problem.
Thus, for operation on the periphery of the image, it is
often necessary to use external data outside the image boundary
according to the number of the taps of the filter used.
Iteration of the subband decomposition also results in
enlarging the area into which the filter extrudes.
In general, the above problems are treated in such a manner
that the image is folded at its periphery according to a certain
6
given rule.
For the Flash Pix method using a plurality of coded data
sets separately provided for respective images of different
resolution levels, the image processing load such as
enlargement or contraction of the image can be reduced, but
the data size is increased to 1.4 times.
For wavelet-transform coding method, data with different
resolution levels can be easily decoded from a single set of
compressed and coded data for an original image size and,
therefore, no-increase in the data size takes place.
When the wavelet-transform coding system utilizes the
method of decomposing an image into tiles and encoding the image
data on a tile-by-tile basis, which is used in the flash-pix
system ( to reduce the processing load by selectively processing
only necessary tiles in case of processing a particular part
of the image) , however, this arises the above-described problem
since filters may stick from the boundary of respective tiles .
In other words, the flash pix system using the JPEG coding
can easily perform coding of each tile owing to the closed
property of coding in each tile, while the wavelet-transform
coding system can not effectively use the above tile-by-tile
coding-and-managing method because the processing causes the
extrusion of filters out of respective tiles.
In addition, the conventional wavelet-transform coding
system must have a memory sufficient for storing an output of
the wavelet-transform portion 31 ( Fig . 3 ) , i . a . , all wavelet
7
transform coefficients as shown in Fig. 5B. Since these
coefficients have the same resolution as that of the original
image, the memory has to possess a large capacity. This
requirement becomes severer when processing a higher
resolution image.
SUMMARY OF THE INVENTION
In view of the above-described problems of the prior art,
the present invention was made to provide a compact hardware
system that realizes effective encoding of images with
different resolutions and effective management of coded data
by tiles by using an improved wavelet-transform technique.
Accordingly, an object of the present invention is to
provide an image coding and decoding system by which the image
is effectively encoded and easily decoded with any resolution
level desired by the user with no increase in volume of coded
data.
This is a great advantage of the present invention system
as compared with the conventional Flash Pix system using the
JPEG coding method, which has an increased amount to 1 . 4 times
of code data to provide a plurality of images having different
resolutions.
Another object of the present invention is to provide an
image-coding and decoding system in which an image is
decomposed into tiles and encoded on a tile-by-tile basis and
the coded tiles can selectively be decoded on the same principle
8
by using the wavelet-transform coding/decoding technique. This
could not be accomplished by the conventional wavelet-transform
coding/decoding system because it is difficult in principle to
apply the wavelet transform to closed tiles of the image.
Another object of the present invention is to provide an image
coding and decoding system that encodes an image on a tile-by-tile
basis and allows the coded image to be partially decoded by
selectively decoding only necessary tiles (without the necessity
of decoding a whole image), thus improving the random access
function of the system.
In order to achieve the above objects, one aspect of the
present invention, which is claimed in the parent application No.
2,303,514, provides an image coding device comprising a tile
decomposition portion for decomposing image data into tiles each
having N pixels x M pixels and outputting the N pixels x M pixels
in the tile as an objective data to be coded for each of the
corresponding tiles; a wavelet-coding portion for extrapolating
predetermined data at a neighboring area of the objective data
input from the tile decomposition portion, decomposing each of the
tiles into subbands and performing separate wavelet-encoding of each
of the tiles; a management information generating portion for
generating management information necessary for independently
decoding any desirable tile from the wavelet-coding portion on a
tile-by-tile basis as well as on a subband-by-subband basis; and
a coded data integrating portion for combining the management
information with the coded data to generate a bit stream, wherein
the management information includes data separately wavelet-encoded
on a tile-by-tile basis and is outputted from the management
information generating portion.
According to a second aspect of the present invention, which
is claimed in the parent application No. 2,303,514 there is
provided an image coding device comprising a tile decomposition
9
portion for decomposing image data into tiles each having N pixels
x M pixels and outputting the N x M pixels in the tile as an
objective data to be coded for each of the corresponding tiles; an
adjacent pixel adding portion for providing an objective tile to be
coded with adjacent pixels necessary for wavelet transformation of
the objective tile when such pixels exist at a periphery thereof;
a wavelet coding portion for extrapolating a predetermined data
when no pixel exists at the periphery of the objective tile
decomposing each of the tiles into subbands and outputting only
wavelet coefficients of the objective tile; a management information
generating portion for generating management inforn~ation necessary
for independently decoding any desirable tile from the wavelet-
coding portion on a tile-by-tile basis as well as on a subband-by-
subband basis; and a coded data integrating portion for combining
the management information to the coded data to generate a bit
stream, wherein the management information includes data separately
wavelet encoded on a tile-by-tile basis and is outputted from the
management generating portion.
According to a third aspect of the present invention, which is
claimed in the parent application No. 2,303,514, there is provided
an image coding device comprising a wavelet-coding portion for
decomposing an image into subbands by extrapolating predetermined
data at a periphery of the image and performing wavelet encoding
of the subbands; a tile composing portion for reconstructing, fr~n
wavelet coefficients inputted from the wavelet coding portion,
separate tiles each being composed of N x M wavelet coefficients
and forming a set to be separately entropy coded; a management
inforniation generating portion for generating management information
necessary for independently decoding any desirable tile from the
wavelet coding portion on a tile-by-tile basis as well as on a
subband-by-subband basis; and a coded data integrating portion fox
composing a sequence of the coded data according to the management
9a
information outputted from the management generating portion and
attaching the management information to the coded data.
According to a fourth aspect of the present invention, which
is claimed in the parent application No. 2,303,514, there is
provided an image decoding device for receiving at its input a bit
stream of coded inforniation including coded information of image
data divided into tiles each containing N pixels x M pixels and
separately wavelet-encoded; tile-position information for specifying
a location of coded data for each tile in the bit stream; and
manat information for managing and identifying s~ubbands
generated by wavelet-encoding of the tiles and for selectively
decoding a coded image corresponding to a necessary tile and
subbands, comprising a management information separarting portion
for separating management information from an input bit stream; a
coded data extracting portion for extracting coded data
corresponding to an objective tile and subbands according to the
management information; a wavelet-decoding portion for conducting
wavelet-decoding of the coded data extracted by the coded data
extracting portion; and a tile-combining portion for combining
decoded images on a tile-by-tile basis into a desired decoded
image.
According to a fifth aspect of the present invention, which is
claimed in the parent application No. 2,303,514, there is provided
an image decoding device for receiving at its input a bit stream
of coded infornlation including coded information of image data
divided into tiles each containing N pixels x M pixels and
separately wavelet-encoded after multiplying each of the tiles data
plus adjacent pixel data by a specified two-dimensional window
function; tile-position information for specifying a location of
each of the tiles in the coded information bit stream; and
management information for managing and identifying subbands
generated when wavelet-encoding of the tiles, and for decoding a
9b
coded image corresponding to a necessary tile and subbands,
comprising a management information separating portion for
separating management information fran the input bit stream; a
coded data extracting portion for extracting coded data part
corresponding to an objective tile and the subbands according to
the management information; a wavelet decoding portion for
conducting wavelet-decoding of the coded data extracted by the
coded data extracting portion; and a tile integrating portion for
arranging wavelet-decoded data at respective places on an original
image and superposing image values at overlaps of neighboring tiles
to integrate the tiles into a desired decoded image.
According to a sixth aspect of the present invention, which is
claimed in the parent application No. 2,303,514, there is provided
an image decoding device for receiving at its input a bit stream
of coded information including coded information of image data
divided into tiles, each of the tiles containing N pixels x M
pixels and separately wavelet-encoded after attaching thereto
adjacent pixels necessary for wavelet-translation of each of the
tiles; tile-position information for specifying the location of the
coded information for each tile in the bit stream; and management
information for managing and identifying subbands generated when
wavelet-encoding of the tiles occurs, and for decoding a coded
image corresponding to a necessary tile and subbands, comprising a
management infornlation separating portion for separating management
information from an input bit stream; a coded data extracting
portion for extracting coded data of an objective tile, tiles
existing around the objective tile and that of the suhbands
according to the management information; a wavelet decoding portion
for conducting wavelet-decoding of the coded data extracted by the
coded data extracting portion; and a tile integrating portion for
arranging wavelet-decoded data at respective places on an original
image and superposing overlapped-part-image values at each overlap
9C
CA 02400487 2003-04-22
of neighboring tiles to integrate the tiles into a desired decoded
image.
According to a seventh aspect of the present invention, which
is claimed in the parent application No. 2,303,514, there is
provided an image decoding device for receiving at its input a bit
stream of coded information including coded information generated
by subband division of image data, tile construction by grouping N
pixel x M pixel wavelet coefficients spatially corresponding to the
tile and entropy-encoding on a tile-by-tile basis; tile-position
information for specifying a location of each of the tiles in the
coded information bit stream; and management information for
managing and identifying subbands generated when wavelet-encoding
of tiles and for selectively decoding a coded image corresponding
to a necessary tile and subbands, comprising a management
information separating portion for separating management information
from the input bit stream; a coded data extracting portion for
extracting coded data of an objective tile and the subbands
according to the management information; a wavelet decoding portion
for conducting wavelet-decoding of the coded data extracted by the
coded data extracting portion; and a wavelet coefficient rearranging
portion for rearranging the wavelet coefficients arranged per tile
inputted in the wavelet decoding, portion to the initial order
before decomposition into tiles.
According to another aspect of the present invention, there is
provided an image decoding device for receiving at its input a bit
stream including coded infozmation of image data divided into tiles
and each separately wavelet-encoded, and management information for
managing the coded information, and for decoding a coded image
corresponding to a necessary tile or a necessary resolution, and
said management information includes information for specifying a
memory location of the coded information corresponding to each tile
or each resolution and information for managing and identifying
9d
CA 02400487 2003-04-22
each tile or each resolution, comprising: an identifying portion
for identifying a memory location of the coded information
corresponding to the tile or the resolution to be decoded with
reference to said management information according to the tile or
the resolution to be coded; a wavelet-decoding portion for
conducting wavelet-decoding of the coded data based on the memory
location of said identified coded information; and a tile-combining
portion for combining the wavelet-decoded images of each tile,
wherein a desired area of image is decoded in a desired resolution.
According to another aspect of the present invention, there is
provided an image decoding device for receiving at its input a bit
stream including coded information of image data divided into tiles
and each separately wavelet-encoded, and management information for
managing the coded information, and for decodsng a coded image
corresponding to a necessary tile, and the management information
includes information far specifying a head location of the coded
information corresponding to each tile andJor information for
managing and identifying each tile, comprising: a wavelet-decoding
portion for conducting wavelet-decoding of the coded data based on
the management information.
According to a further aspect of the present invention, there
is provided an image encoding device, comprising: a tile-dividing
portion for dividing image data into tiles; a wavelet-encoding
portion for conducting a wavelet-encoding of each tile separately
to generate a coded information; a management information generating
portion for generating management information to manage said coded
information; and a coded data integrating portion for integrating
said management information and said coded info~nation to generate
a bit stream, wherein the management information includes
information for specifying a memory location of the coded.
information Corresponding to each tile or each resolution and
9e
CA 02400487 2003-04-22
information for managing and identifying each tile or each resolution.
According to another aspect of the present invention, there is
provided an image encoding device, comprising: a title-dividing
portion for dividing image data into tiles; a wavelet-encoding
portion for conducting a wavelet-encoding of each tile separately
to generate a coded information; a management information generating
portion for generating management information to manage to Coded
information; and a coded data integrating portion for integrating
the management information and the coded information to generate a
bit stream, wherein the management information includes information
for specifying a head location of the coded information
corresponding to each tile and/or information for managing and
identifying each tile.
According to another aspect of the present invention, which is
claimed in Canadian Divisional Application No. 2,400,496,
there is provided an image decoding method of receiving at its
input a bit stream including coded information of image data
divided into tiles each containing N pixels x M pixels and each
separately wavelet-encoded, and management information including
information for specifying a memory location of the coded
information corresponding to each tile or each resolution and
information for managing and identifying each tile or each
resolution, and of decoding a coded image corresponding to a
necessary tile and a necessary resolution, comprising the steps of:
separating the management information from an input bit stream;
setting a tile or a resolution to be decoded according to a user's
instructions; identifying a memory location of the coded information
corresponding to the tile or the resolution to be decoded with
reference to said management information according to said setting
step; extracting a portion of coded information from the bit
stream, the portion of the coded information corresponding to the
set tile or the resolution based on the merry location of said
9f
CA 02400487 2003-04-22
identified coded information; conducting wavelet-decoding of the
coded data extracted by said coded data extracting step; and
combining the wavelet-decoded images on a tile-by-tile basis,
wherein a desired area of image is decoded in a desired resolution
according to a user's instructions.
According to another aspect of the present invention, which is
claimed in Canadian Divisional Application No. 2,400,496,
there is provided an image decoding method. of receiving at its
input a bit stream including coded information of image data
divided into tiles each containing N pixels x M pixels and each
separately wavelet-encoded, and management information including
information for specifying a memory location of the coded
information corresponding to each tile or each resolution and
information for managing and identifying each tile or each
resolution, and of decoding a coded image corresponding to a
necessary tile and a necessary resolution, comprising the steps of:
reading management information of said management information, the
management information being added to said coded information;
reading management information of said management information, the
management information including information for specifying the
memory location of the coded information corresponding to said each
tile or each resolution and being disposed at a location which is
independent of said coded information; setting a tile or a
resolution to be decoded according to a user's instructions;
identifying a memory location of the coded information corresponding
to the tile or the resolution to be decoded with reference to said
management information according to said setting step; extracting
a portion of coded information from the bit stream, the portion of
the coded information corresponding to the set tile or the
resolution based on the memory location of said identified coded
information; conducting wavelet-decoding of the coded data extracted
by said coded data extracting step; and combining the wavelet-
9g
o CA 02400487 2003-04-22
decoded images on a tile-by-tile basis, wherein a desired area of
image is decoded in a desired resolution according to a user's
instructions.
According to a further aspect of the present invention, which
is claimed in Canadian Divisional Application No.
2,400,496, there is provided an image decoding method of receiving
at its input a bit stream including coded information of image data
divided into tiles each containing N pixels x M pixels and each
separately wavelet-encoded, and management information including
information for specifying a head portion of the coded information
corresponding to each tile and information for managing and
identifying each tile, and for decoding a coded image corresponding
to a necessary tile, comprising the steps of: separating the
management information from an input bit stream; extracting a
portion of the coded information from the bit stream, the portion
of the coded information corresponding to a given tile based on
said management information; and conducting wavelet-decoding of the
coded data extracted by said coded data extracting step.
BRIEF DESCRIPTION OF THE DRAWINGS
Figure 1 is a view fox explaining how to reduce an image in
size and decompose each image into tiles by a prior art system.
Figure 2 shows an exemplary coding device for encoding an image
1 shown in Fig. 1.
Figure 3 is a basic block diagram of a wavelet-coding portion.
Figure 4 is a detailed block diagram of a wavelet transform
portion.
Figure 5 is a view for explaining a correlation between an
original image and wavelet-transformed data..
Figure 6 is a basic block diagram of a wavelet decoding
portion.
Figure 7 is a view for explaining vertical and horizontal
9h
filters for wavelet-transform of an original image.
Figure 8 is a block diagram of an image coding device
according to an embodiment 1 of the present invention.
Figure 9 is view for explaining the operation of an image
coding device according to the embodiment 1 of the present
invention.
Figure IO shows an exemplified bit stream in an image coding
device according to the embodiment 1 of the present invention.
Figure 11 shows another exemplified bit stream in an image
coding device~according to the embodiment 1 of the present
invention.
Figure 12 is view for explaining the operation of an image
coding device that is an embodiment 2 of the present invention.
Figure 13 is a block diagram of an image decoding device
according to an embodiment.3 of the present invention.
Figure 14 a.s view for explaining the operation of an image
decoding device according to an embodiment 4 of the present
invention.
Figure 15 is a block diagram of an image coding device
according to an embodiment 5 of the present invention.
Figure 16 is view for explaining the operation of an image
coding device according to the embodiment 5 of the present
invention.
Figure 17 is view for explaining the operation of an image
coding device according to an embodiment 6 of the present
invention.
Figure 18 is a block diagram of an image coding device
according to an embodiment 7 of the present invention.
Figure 19 is view for explaining the operation of an image
coding device according to the embodiment 7 of the present
invention.
Figure 20 is a block diagram of an image decoding device
according to an embodiment 8 of the present invention.
Figure 21 is a block diagram of an image decoding. device
according to an embodiment 9 of the present invention.
Figure 22'is a block diagram of an image coding device
according to an embodiment 10 of the present invention, with
a view for explaining the operation of the same device.
Figure 23 is a block diagram of an image decoding device
according to an embodiment 11 of the present invention, with
a view for explaining the operation of the same device.
Figure 24 is a block diagram of an exemplary image coding
device according to an embodiment 12 of the present invention.
Figure 25 is a block diagram of another exemplary image
coding device according to the embodiment l2 of the present
invention.
Figure 26 is a block diagram of another exemplary image
coding device according to the embodiment 12 of the present
invention.
Figure 27 is a block diagram of an exemplary image decoding
device according to an embodiment 13 of the present invention.
Figure 28 is a block diagram of another exemplary image
11
decoding device according to the embodiment 13 of the present
invention.
Figure 29 is a block diagram of another exemplary image
decoding device according to the embodiment 13 of the present
invention.
Figure 30 is a block diagram of an exemplary image coding
device according to the embodiment 14 of the present invention.
Figure 31 shows an exemplified bit stream in'an image coding
device according to the embodiment 14 of the present invention.
Figure 32~is a block diagram of another exemplary image
coding device according to the embodiment 14 of the present
invention.
Figure 33 is a block diagram of another exemplary image
coding device according to the embodiment 14 of the present
invention.
Figure 34 is a block diagram of another exemplary image
coding device according to the embodiment 14 of the present
invention.
Figure 35 is a block diagram of an image decoding device
according to an embodiment 15 of the present invention.
Figure 36 is a block diagram of an exemplary image coding
device according to an embodiment 16 of the present invention,
with a view for explaining the operation of the same device.
Figure 37 is a block diagram of another exemplary image
coding device according to the embodiment 16 of the present
invention.
12
Figure 38 shows an image decoding device according to an
embodiment 17 of the present invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
Figure 8 is a block diagram showing the construction of
an image coding device that is an embodiment 1 of the present
invention.
Image data of an original image as shown in Fig. 9A is
decomposed by a tile decomposition portion 101 into tiles each
of predetermined N pixels by M pixels. The decomposed image
is shown in Fig. 9B. The tile decomposition portion 101 outputs
N pixels by M pixels image in a tile as corresponding data to
each tile.
Further processing will be described by way of example on
a tile "i" in Fig. 9B. Image data of the tile "i" is divided
by a wavelet transform portion 102 into subbands.
Data at the periphery of a tile is extrapolated when
dividing the tile portion near its boundary into subbands . For
example, as shown in Fig. 7B, an area 92 covered by a wavelet
transform filter exists out of a tile. In this case, it is needed
to add data at the periphery of the tile. The wavelet transform
portion 102 therefore extrapolates data at the periphery of
each tile and divides the tile into subbands.
The data extrapolation is achieved for example by
generating a mirror image by outwardly folding an internal
image of the tile as shown in Fig. 9C. A quantizing portion
13
103 quantizes wavelet transform coefficients and an entropy
coding portion 104 performs entropy coding of the coefficients
to obtain coded data of the tile "1".
The entropy coding can be achieved by using a known Huffman
coding method or arithmetic coding method. The wavelet
transform portion 102, quantizing portion 103 and entropy
coding portion 104 composes a so-called wavelet transform
coding portion 105.
On the other hand, a management information generating
portion 106 generates information for identifying and managing
tiles and subbands by using information on spatial locations
of each tile from the tile decomposition portion 101 and
information on each subband from the wavelet transform coding
portion 105. The management information is utilized by a coded
data integration portion 10'x.
Using the management information from a management
information generating portion 106 , the coded data integration
portion 107 arranges and integrates information on the coded
data from the entropy coding portion and adds the management
information to a bit stream to generate coded data.
Management of the coded data according the tiles and
subbands is needed for achieving decoding of a coded image at
different resolution levels as shown in Fig. 1 or a particular
tile or tiles of the coded image.
Figure 10 shows an example of a bit stream of coded data
produced in the above-described manner. The bit stream is
14
composed of a header for managing information on a whole bit
stream and information on each tile. Information for each tile
consists of a tile header for managing the tile information
and coded information representing a tile image encoded by the
wavelet transform coding portion I05.
The tile header includes information on bit positions
corresponding to respective subbands. A bit sequence
corresponding to necessary one of the subbands can be found
by accessing this information.
The structure of bit streams used in the system of the
present invention is not limited to that shown in Fig. 10. For
example, a sequence ( I ) of Fig. 11 is similar to the sequence
of Fig. 10, while a sequence (II) of Fig. 1l has the form in
which each subband for a tile is separated and rearranged with
a tile header added thereto .~ The latter sequence ( I I ) allows
the system to quickly ,reproduce a desirable contracted image
by accessing only necessary tile or tiles in the sequence.
An image coding device according another embodiment 2 of
the present invention will be described as follows . The image
coding device of the embodiment 2 is similar in construction
to the embodiment 1 shown in Fig. 8 but differs from the
embodiment 1 described above with the figure 8 by the operation
of the tile decomposition portion 101, which will be described
below with reference to Fig. 12.
While the tile decomposition portion 101 of the embodiment
1 decomposes an image into tiles each of N pixels by M pixels
Z5
and outputs only image data within each tile to the wavelet
transform portion I02, the tile decomposition portion 101 of
embodiment 2 outputs image data obtained by multiplying the
original image by a suitable window function.
For example, in the case of extracting a tile "ij", the output
of the tile decomposition portion 101 is a result of multiplying
the original image data by a window function FXi in the
horizontal direction, and by a window function Fyj in the
vertical direction. i denotes a horizontal tile number and j
denotes a vertical tile number.
This means that the output of the tile decomposition
portion I01 represents a weighted result of multiplying a
shaded image portion (Fig. 12) by a weight corresponding to
window functions . Window functions are such that a total of
functions over a whole area is equal to 1. Window functions
satisfying the following conditions are used.
EFXi(x)=1 (OSxSw)
EFYj(Y)=1 (05YSh)
where, w is the width of the original image, h is the height
of the original image, x and y are the axes of abscissa and
ordinate, respectively, with the origin at the top right corner
of the original image.
A total of the functions FXi(x) is taken for i and FYj(y)is
taken for j . In Fig . 12 , FXi_1, FXi, FXYi , FY j , FY j,1 are exemplary
functions satisfying the above conditions.
In consequence of the extraction of data by applying window
16
functions, the output of the tile decomposition portion 101
includes pixels of a tile ij plus peripheral pixels weighted
with the window function values.
An image decoding device for decoding coded data from the
image coding device of the embodiment 1 will be now described
as an embodiment 3 of the present invention. Figure 13 is a
block diagram of the image decoding device according to the
embodiment 3.
The image decoding device receives coded data from the
image coding device described as the embodiment 1 of the present
invention. A management data separating portion 111 takes out
information for managing tiles and subbands from the received
coded data.
A coded data extracting portion 112 selectively extracts
coded data of necessary tile and subbands according to the
user' s request . In the exemplary bit stream shown in Fig . 10 ,
the management information is found in the header and the tile
header.
The extracted coded information is entropy-decoded by an
entropy decoding portion 113 and inversely quantized by an
inverse quantizing portion 114 to produce wavelet-transform
coefficients corresponding to the tile to be decoded.
The wavelet-transform coefficients are inversely
transformed by an inverse wavelet transform portion 115 to
produce a decoded image of the objective tile. The
entropy-decoding portion 113, inverse quantizing portion 114
17
and inverse wavelet-transform portion 115 compose a so-called
wavelet-transform decoding portion 116.
A tile combining portion 117 combines together decoded
tiles according to the tile managing information to generate
a decoded image of the desired area or at a desired resolution.
The decoding process with the bit stream shown in Fig. 10
is as follows. To decode a low-resolution entire image (all
tiles ) , coded data ( 1-a, 2-a, ..., i-a, ...) , which correspond to
low-resolution subbands, are decoded in order in respective
tile by the wavelet-transform decoding portion 116 according
to the tile with referring to subband information included in
each tile header.
The low-resolution decoded tiles are then combined by the
tile-combining portion 117, thereby a whole low-resolution
image is reproduced.
From the low-resolution decoded image, a particular tile
"i" can also be reproduced in an enlarged scale with the highest
resolution by decoding all the coded information of the i-
th tile which correspond to the tile image "i".
Namely, coded information i-b extracted and decoded
together with already extracted coded information 1-a to obtain
the desired decoded image. It is, of course, possible to
reproduce a high-resolution decoded image of all areas by
decoding all coded information (all tiles including all
subbands).
Thus, the image decoding device can easily decode any
18
resolution image and/or any tile (partial) image can be easily
decoded according to the user's request.
An image decoding device according to another embodiment
4 of the present invention is as follows:
Coded data is input from the image coding device according
to the embodiment 2 of the present invention. This image
decoding device is similar in construction to the embodiment
3 shown in Fig . 13 and differs from the latter by the operation
of the tile-combining portion 117, which will be described
below with reference to Fig. 14.
In the image coding device according to the embodiment 2 ,
pixels of each tile have been encoded together with pixels at
its periphery. Therefore, data of a tile decoded in a
wavelet-transform decoding portion 116 of this image decoding
device is larger than an actual tile.
In Fig . 14 , a tile is composed of 2 x 2 pixels and the data
size of a decoded tile is of 4 x 4 pixels. In this case, decoded
data of a tile ij has an area shaded in Fig. 14, which overlaps
neighbors each by one pixel width.
The tile-combining portion 117 determines a pixel value
for each overlap by adding together decoded data thereat when
linking the decoded tiles. For example, the value of a pixel
"a" in Fig. 14 is calculated as follows:
a(i-l,j-1)+a(i,j-1)+a(i-l, j)+a(i,j)
where a(i,j) represents decoded data of the tile ij at the
position of its pixel °a".
19
An image coding device according to another embodiment 5
of the present invention will be described with reference to
Fig. 15 showing its construction.
This image coding device differs from the image coding
device (embodiment 1) of Fig. 8 by the fact that it does not
unconditionally conduct extrapolation of data at the periphery
of an objective tile and utilizes another tile adjacent to the
tile if such exists.
Like the embodiment 1, this image coding device decomposes
an original image into tiles as shown in Fig. 16A at its tile
decomposition portion 121 shown in Fig. 15. The image coding
device further processes a tile "i" of the image as follows
In a wavelet-transform coding portion 123, image data of the
tile "i" is wavelet-transformed through a wavelet-transform
filter. In this case, if the .filter extrudes from the tile "i"
into neighboring tiles and covers part of pixels contained in
the neighbors , image data of those pixels in the neighbors are
also wavelet-transformed together with the image data of the
objective tile "i" by the filter.
Referring to Fig. 16, the objective tile "i" of Fig. 16A
is extended by adding necessary shaded parts of neighboring
tiles a-h as shown in Fig. 16B and then wavelet-transformed.
An adjacent pixel adding portion 122 realizes the above
process by recognizing neighboring tiles around the objective
tile according to the tile decomposition information and by
adding necessary pixels if the neighbors exist.
In the shown case, the adjacent pixel adding portion 122
adds to the objective tile "1" all neighboring pixels existing
around there and outputs an increased volume of image data to
a wavelet transform portion 123. The wavelet transform portion
123 must therefore transform a larger image area data as
compared with the wavelet transform portion 102 of the
embodiment 1 that transforms a single tile image.
With an increased image data to be transformed, the image
coding device requires an enlarged working area for processing
the data, resulting in increasing the cost of the device and
decreasing the operating speed. Therefore, it is desirable to
use another mode for reducing the size of data to be transformed
by the wavelet-transform portion.
This mode is such that the additional adjacent area to be
added by the peripheral pixel adding portion 122 to an objective
tile area is limited to one direction "x" or "y" as shown in
Fig. 16C or 16D to reduce the image data to be input to the
wavelet transform portion 123.
For example, with reference to Fig. 16C, an objective tile is
provided with necessary pixels from upper and lower neighboring
tiles if they exist. The right and left sides of the objective
tile are treated by generating mirror images by folding the
inside image of the tile. The case of Fig. 16D differs from the above
case of Fig. 16C by adding pixels from the right and left
neighbors and generating mirror images for its top and bottom
sides.
21
There are two alternative methods for performing the
wavelet transform of the image data of the objective tile. One
method is to recursively discompose any of the tile of Figs.
16B, 16C and 16D. The other method is to selectively apply any
of the above three pixel-addition modes (Figs. 16B, 16C and
16D) for each subband.
The wavelet transform portion 123 generates wavelet
transform coefficients of pixels included within the ob jective
tile "i" using only for calculation purpose of wavelet
transform coefficients of the pixels added thereto by the
peripheral pixel adding portion 122.
A quantizing portion 124 quantizes the wavelet transform
coefficients and an entropy coding portion 125 performs entropy
encoding of the quantized coefficients to obtain coded
information of the objective tile "i". The wavelet transform
portion 123, quantizing portion124 and entropy-coding portion
125 composes a so-called wavelet transform coding portion 126 .
On the other hand, a management information generating
portion 127 receives spatial-tile-position information from
the tile decomposition portion 121 and subband information from
the wavelet transform coding portion 126 and generates
management information for managing and identifying tiles and
subbands . The management information is used by a coded-data
integrating portion 128.
The coded data integrating portion 128 rearranges and
integrates coded information outputted from the entropy coding
22
portion 125 according to the management information outputted
from the management portion 127 and then adds the management
information to a bit stream to generate a final coded data as
shown for example in Fig. 10.
An image coding device according to still another
embodiment 6 of the present invention will be described below:
This image coding device is similar in construction to the
device of the embodiment 5 described above with reference to
Fig. 15 but differs from the latter only by the operation of
its peripheral-pixel adding portion 122. The operation of the
peripheral-pixel adding potion 122 is described below with
reference to Fig. 17.
An objective tile "1" in Fig. 17 is now processed by way
of example as follows:
In the embodiment 5, the_peripheral-pixel-adding portion
122 added to a tile "i" all pixels necessary for calculating
wavelet-transform coefficients for pixels in the objective
tile, that is, pixels in areas covered by a filter extending
from the objective tile. The adjacent pixel areas are shown
as shaded in Fig. 17.
Since distant pixels have a small effect on wavelet
transform coefficients in a tile "i", the embodiment 6 adds
a result of multiplying peripheral pixels by a suitable
weighing function to the tile "i" to reduce the number of pixels
to be attached, i.e., lighten the computation work load.
The weighting function is 1 for each pixel near the tile
23
"i" and has a distance-depending value approaching to zero as
a distance from the tile "i" increases. In Fig. 17, there is
an example of a weighting function. Pixels multiplied by the
weighting function and actually added to the objective tile
compose an effective pixel area mesh-dotted in Fig. 17. A
peripheral pixel area shown as shaded only in Fig. 17 is
necessary for wavelet transform calculation but is not added
because its weighted value is zero.
Alternatively, a stepwise weighting function may be
applied, which is given 1 for each pixel within a specified
distance from the tile °i" and 0 for all pixels existing over
the specified distance.
Another image coding device is- described as an embodiment
7 of the present invention. Figure 18 is a block diagram of
an image coding device according to the embodiment 7.
This image coding device differs from the embodiment 1 of
Fig. 8 and the embodiment 5 of fig. 5 by the fact that an original
image is entirely wavelet-transformed.by a wavelet transform
portion 131 and, then, wavelet transform coefficients
outputted from the wavelet transform portion 131 are rearranged
per tile to compose respective tiles.
In Fig. 18, an original image before tiling is
wavelet-transformed by a wavelet transform portion 131. A tile
composing portion 132 composes tiles by rearranging wavelet
transform coefficients so that a tile is composed of
coefficients spatially matching the same tile.
24
Figure 19A shows an example of subbands obtained by wavelet
transform portion 131. In the shown case, a coefficient b0 in
the lowest frequency subband spatially correlates with other
subband coefficients b1, b2, b3, b4, b5, b6, b7, b8 and b9.
Where bl-b3 consist each of 1x1 coefficient, b4-b6 consist
each of 2x2 coefficients and b7-b9 consist each of 4x4
coefficients. These coefficients b0-b9 are taken out of the
respective subbands and then arranged to compose a single tile
as shown in Fig. 198. Likewise, all ether wavelet transform
coefficients are arranged to compose respective tiles. This
results in obtaining the same result as in the embodiment 5
whereby an original image is first decomposed into portion
tiles and then wavelet transformed.
b0 is not necessarily a single coefficient, but it may be
a block composed of k x 1 coefficient.
In this case, bl-b3 consist each of k x l, b4-b6 consist
each of 2k x 21 and b7-b9 consist each of 4k x 41 coefficients.
Wavelet transform coefficients organized per tile are
outputted from the wavelet transform portion 132. They are
quantized by a quantizing portion 133 and entropy-encoded by
an entropy-coding portion 134, thus coded information is
generated.
On the other hand, a management-information generating
portion 136 generates management information necessary for
managing and identifying tiles and subbands using spatial-
tile-location information from the tile composing portion 132
and subband-location information from the wavelet transform
coding portion 135. The management information is used by a
coded-information integrating portion 137.
The coded information integrating portion 137 receives the
management information from the management information
generating portion 136 and the coded information from the
entropy-coding portion 134 and it arranges and combines the
entropy coded information and adds management -information in
a bit stream of the coded data, thus generating finally coded
data as shown-in Fig. 10.
Although the tile-composing portion 132 is installed
before the quantizing portion 133 in this embodiment, it is
not limited to this arrangement and may be placed after the
quantizing portion 133.
An image decoding device, for decoding data encoded by any
one of the above-described image coding devices (embodiment
to 7 ) is now described below as an embodiment 8 of the present
invention. Figure 20 is a block diagram showing the
construction of the image decoding device according to the
embodiment 8. The decoding device receives coded data encoded
by any one of the image coding devices described above as
embodiments 5 to 7.
Referring to Fig. 20, the image decoding device separately
takes out tile-decomposition management information and
subband-management information from the input coded-data
stream by a management information separating portion 141 and
26
selectively extracts a necessary Bart of the coded information
meeting the user's demand by a coded data extracting portion
142 according to the management information. Namely, the coded
data corresponding to a necessary objective tiles) having a
necessary resolution is extracted by the coded data extracting
portion 142.
The extracted coded information is entropy-decoded by an
entropy decoding portion 143 and inversely quantized by an
inverse quantizing portion 144. Thus, wavelet transform
coefficients corresponding to an objective tile to be decoded
are now obtained.
The wavelet transform coefficients inversely transformed
by an inverse wavelet transform portion 145, thereby a decoded
image including peripheral pixels data is produced. The entropy
coding portion 143, the inverse,quantizing portion 144 and the
inverse wavelet transform portion 145 compose a so-called
inverse wavelet transform decoding portion 146.
A tile combining portion 147 integrates groups of the
decoded tiles according to the management information from the
management information-separating portion 141. In this case,
a completely decoded image is reproduced with overlaps of
decoded tile images at each spatially overlapped portion.
Namely, the embodiment 2 described above with reference
to Fig. 12 performed wavelet transform of each tile with
adjacent pixels attached thereto. The embodiment 5 uses
adjacent pixels in performing wavelet transform of each tile
27
as shown in Fig . 16B . Likewise , the embodiment 6 described with
reference to Fig. 17 also uses peripheral pixels in wavelet
transform of each tile.
In the image coding device according to the embodiment 7 ,
the process using adjacent pixels is not clearly described but
the wavelet transform of a whole original image has been done
including the processing theoretically equivalent to that made
in the embodiment 5.
Therefore, data of peripheral pixels is produced when each
tile image is decoded by the wavelet transform decoding portion
146 in Fig. 20 and the decoded adjacent pixels are superposed
on respective neighboring tiles by the tile combining portion
147. The superposition of one pixel on another is achieved by
additive operation on the pixels.
Another image decoding,device is described below as an
embodiment 9 of the present invention. Like the above
embodiment 8, the input to this embodiment 9 is coded data
encoded by any one of the image coding devices being the
embodiments 5 to 7. Figure 21 is a block diagram showing the
construction of the image coding device according to the
embodiment 9.
Referring to Fig. 21, a management information separating
portion 151 separately takes out tile-division management
information and subband management information from the input
coded-data stream, and a coded data extracting portion 152
selectively extracts a necessary part of the coded information
28
meeting the user's demand according to the management
information. Namely, the coded data corresponding to a
necessary objective tiles) having a necessary resolution is
extracted by the coded data extracted portion 152.
The extracted coded information is entropy-decoded for
each tile by an entropy-decoding portion 153 and inversely
quantized by an inverse quantizing portion 154. Wavelet
transform coefficients corresponding to an objective tile to
be decoded are thus obtained. A wavelet transform coefficient
rearranging portion 155 rearranges the wavelet transform
coefficients into the state in which they were placed before
tile-by-tile arrangement.
Namely, the wavelet transform coefficients divided per
tile as shown in Fig. 19B are rearranged in the state shown
in Fig. 19A. After completion of processing on all tiles, all
wavelet transform coefficients of Fig. 19A are obtained.
The rearranged wavelet transform-coefficients can be
decoded at a time by inverse transformation. Namely, the
coefficients are inversely transformed by an inverse wavelet
transform portion 156, whereby a whole decoded image is
reproduced.
The entropy coding portion 153, the inverse quantizing
portion 154 and the inverse wavelet transform portion.156
compose a so-called inverse wavelet transform decoding portion
157. Although the wavelet transform coefficient rearranging
portion 155 is installed after the inverse quantizing portion
29
154 in this embodiment, it is not limited to this arrangement
and may be placed before the inverse quantizing portion 154.
An image coding device is described below as an embodiment
of the present invention. Figure 22E is a block diagram of
a portion of this embodiment, which responds to the wavelet
transform portion (102 in Fig. 8, 123 in Fig. 15) of the image
coding devices according to the embodiments 1, 2, 5 and 6.
Referring to Fig. 22E, a memory 162 is used for storing
wavelet transform coefficients divided into subbands by a
wavelet transform portion 161. In this case, the memory 162
stores only wavelet transform coefficients corresponding to
a tile being currently processed by the wavelet transform
portion 161. The processed data are transferred to a quantizing
portion (103 in Fig. 8, 124 in Fig. 15) following the wavelet
transform portion 161.
Therefore, the memory 162 has no need to store all data
for a whole image and is sufficient to store such an amount
of data necessary fox processing only one tile.
Namely, if wavelet-transformation without tile
decomposition is applied to a whole image as shown in Fig. 22A,
it is necessary to store all wavelet transform coefficients
(Fig. 22B) outputted from the wavelet transform portion 161.
In contrast to the above, the decomposition of an image into
tiles as shown in Fig. 22C enables the coding device to use
a memory for storing only wavelet transform coefficients
corresponding to a small image of Fig. 22D, thus realizing a
considerable saving of the memory capacity.
The same effect can be realized in an image decoding device.
An image decoding device is described below as another
embodiment 11 of tha present invention. Figure 23E is a block
diagram, which corresponds to the inverse wavelet transform
portion ( 115 in Fig . 13 , 145 in Fig . 20 ) of the image decoding
devices described before as the embodiments 3, 4 and 8.
Referring to Fig. 23E, a memory 171 stores wavelet
transform coefficients necessary for decoding one tile and an
inverse wavelet transform portion 172 performs the composition
of subbands.
An image that must be decoded is assumed to be that shown
in Fig. 23B. When performing the wavelet transform of the image
without decomposition into tiles, it is necessary to store all
wavelet transform coefficients as shown in Fig. 23A. On the
contrary, when decoding an image decomposed into tiles as shown
in Fig. 23D, the image decoding device can operate using a
memory 171 storing the limited number of wavelet transform
coefficients as shown in Fig. 23C. The necessary memory
capacity can be considerably saved.
All the above-described embodiments can be provided with
a plurality of subband-decomposition filters that are
adaptively switched over one another to use in the process of
wavelet transform coding.
The subband decomposition filters mean low-pass filters
and high-pass filters for decomposing an image into subbands
31
as described before for the prior art devices. The subband
decomposition.process is iterated for wavelet transformation.
Filters to be used for this purpose are of various types having
different numbers of taps and different coefficient values.
Accordingly, it is desirable to selectively apply suitable
one of filters to each subband-decomposition because this
enables the coding device to change a necessary amount of
adjacent pixels for an objective image by applying a suitable
filter for a current subband. Optimal wavelet transformation
of an image may be achieved by finding a reasonable compromise
between the processing data amount and the image quality.
In image decoding devices responding to such image coding
devices, each of subband composition filters responding to
respective subband decomposition, filters used for wavelet
transformation are selectively used for each of subbands to
be combined through inverse wavelet transformation.
An image coding device is described below as another
embodiment 12 of the present invention . hn the embodiment 12 ,
an input image can be encoded by one of a plurality of
predetermined coding methods.
Figure 24 is a block diagram showing an exemplary image
coding device according to the embodiment 12. This embodiment
12 performs the image coding by switching the coding mode from
the method of the embodiment 1 to the method of the embodiment
7 or vice versa.
Referring to Fig. 24, a tile wavelet-coding portion 201
32
performs wavelet encoding of the input image on a tile-by-
tile basis and outputs coded information. This tile wavelet
coding portion 201 also outputs tile-decomposition
information, subband information and flag information.
A tile-management-information generating portion 203
receives the tile-decomposition information, the subband
information and the flag information, and it prepares and
provides management information including a combination of the
above inputs. A coded data integration portion 107 outputs
encoded data which combines the coded information and the
management information.
In the tile wavelet-coding portion 201, an input original
image is decomposed into tiles by the tile decomposition
portion 101 and the.decomposed image (tiles) is input to a
terminal 0 of a first switch .204. The original image is input
to a terminal 1 of the first switch 204. Either of two images
through the switch 204 is input to a wavelet transform coding
portion 207.
The wavelet transform coding portion 207 performs wavelet
encoding of the input image. The output of a first wavelet
transform portion 208 is input to a quantizing portion 103
through a second switch 205 or to the quantizing portion 103
through a tile-composing portion 132.
The operation of the first wavelet transform portion 208
is similar to the wavelet transform portion 102 of the
embodiment 1 described with reference to Fig . 8 . So , the portion
33
is not further described.
A flag generating portion 202 outputs a flag for selecting
the encoding method of the embodiment 1 or the encoding method
of the embodiment 7 and, at the same time, controls the first
switch 204, second switch 205 and third switch 206.
When the switches 204 , 205 , 206 are connected to terminals
0 , the coding device performs the coding operation in the same
way as the embodiment 1 does . With the switches connected to
terminals 1, the coding device conducts the coding operation
in the same way as the embodiment 7 does.
The operation of a tile-composing portion 132 is the same
as that of the embodiment 7 described before with reference
to Fig. 18. Further description is omitted.
As described above, the present embodiment can encode an
input image on a tile-by-tile basis and selectively switches
the coding system to the method of the embodiment 1 featured
by simple image-by-image processing or the method of the
embodiment 7 featured by coding of each tile with no distortion
at the boundary thereof.
Figure 25 is a block diagram of another exemplary image
coding device according to the embodiment 12. In this coding
device, coding can be conducted by selectively applying the
method of the embodiment 1 and the method of the embodiment
5.
Referring to Fig. 25, the image coding device differs from
the former type by omitting the tile composing portion 132 (Fig.
34
24) relating to the embodiment 7 and adding an adjacent pixel
adding portion 122 relating to the embodiment 5 and a second
wavelet transform portion 305 with a selector switch. Since
the operation of the components of this coding device except
tile wavelet-transform coding portions 301 and 302 (Fig. 25)
are similar to those of the image coding device of Fig. 24,
so further description is omitted.
The wavelet transform coding portion 302 performs wavelet
coding of an input image and outputs coded information. This
device has two inputs: one is connected to a first wavelet
transforming portion 208 and the other is connected to a second
wavelet transforming portion 305.
When an image is input to the first wavelet transform coding
portion 208, the wavelet transform coding portion 302 performs
the same operation as the wavelet transform coding portion 207
of Fig . 24 does . When an image is input to the second wavelet
transform portion 305, the wavelet transform coding portion
302 performs the same operation as the wavelet transform coding
portion 126 of Fig. 15 does, since the operation of the first
wavelet transform portion 305 is similar to that of the wavelet
transform portion 123 of Fig. 15.
In tile wavelet coding portion 301, the input image is
decomposed into tiles and transferred to a first switch 303.
On the other hand, the decomposed tile images with adjacent
pixels are inputted to a second switch 304. A flag-generating
portion 306 selects the use of the first wavelet-transform
portion 208 or the second wavelet-transform portion 305 in the
wavelet transform coding portion and outputs a flag indicating
the selection made.
At the same time, the above selection causes the first
switch 303 or the second switch 304 to turn ON. Once the first
switch was turned ON, the decomposed image is inputted to the
first wavelet transform portion 208 whereby the same coding
process as made in the embodiment 1 is performed. Once the
second switch 304 was selected, the image decomposed into tiles
with peripheral pixels is inputted to the second wavelet-
transform portion 305 whereby the coding process of the
embodiment 5 is performed.
Thus, the image coding device can process an input image
on a tile-by-tile basis and can encode the image by selectively
applying the simple coding method of the embodiment 1 or the
tile-boundary distortionless coding method of the embodiment
5, (or the coding method of the embodiment 5, which can
encode each tile without distorting at least upper and lower
or (or/and) left and right of the boundary thereof).
Another exemplary image coding device according to the
embodiment 12 is shown in Fig. 26, which is capable of
selectively applying three different coding modes: methods of
the embodiments 1, 5 and 7.
As shown in Fig. 26, this image coding device differs in
construction from the image coding device of Fig. 25 by
including a tile composing portion 132 and switching circuitry
for realizing the coding mode of the embodiment 7 . The operation
36
of this device excepting a tile wavelet transform coding
portion 401 and a wavelet transform coding portion 407 is
similar to that of the device of Fig. 24, so further description
of the portions is omitted.
The wavelet transform coding portion 407 performs wavelet
encoding of an input image and outputs coded information of
the image. The output of a first wavelet transform portion 308
is inputted to a quantizing portion 103 through a third switch
405 or further through a tile-composing portion 132 . The output
of a second wavelet-transform portion 305 is inputted directly
to the quantizing portion 103.
In the tile wavelet-coding portion 401, the input image
is inputted directly to a terminal 0 of the first switch 403.
Alternatively, it is decomposed into tiles and then inputted
to a terminal 1 of the first. switch 403, or it is decomposed
into tiles each including necessary peripheral pixels and then
inputted to a terminal 2 of the switch 403.
These images are transferred to a first wavelet-
transforming portion 308 or a second wavelet-transforming
portion 305 through the second switch 404. The image data is
quantized in a quantizing portion 103 and encoded in an
entropy-coding portion 104 wherefrom coded information is
outputted.
A flag generating portion 402 controls the first, second,
third and fourth switches 403, 404, 405, 406 to selectively
switch the coding modes 0, 1 and 2. The mode numbers are
37
indicated at terminals of the switches 403, 404, 405 and 406
respectively in Fig. 26.
When the first switch 403 is connected to the terminal 0,
all remaining switches 404, 405 and 406 are also connected to
their terminals 0. With the switches 403-406 connected to the
terminals 0 , the image coding device encodes the input image
by applying the coding mode of the embodiment 7.
When the switches 403-406 are all connected to their
terminals 1, the image coding device encodes the~input image
by applying the coding mode of the embodiment I . When the first,
second and fourth- switches 403 , 404 , 406 are connected to their
terminal 2, the image coding device encodes the input image
by using the~coding mode of the embodiment 5.
Thus, the image coding device can process an input image
on a tile-by-tile - basis and can also encode the image by
selectively applying one of three coding modes: the simple
tile-image-coding method of the embodiment 1, the tile-
boundary distortionless coding method of the embodiment 5 or
7, (or the coding method of the embodiment 5, which can
encode each tile without distorting at least upper and lower
or (or/and) left and right of the boundary thereof).
An image decoding device according to another embodiment
13 of the present invention is described below. The input to
this device is coded data encoded by the image coding device
according to the embodiment 12 of the present invention. The
input data is decoded by this device by applying one of
predetermined modes of decoding.
38
Fig. 27 is a block diagram of an exemplary image decoding
device according to the embodiment 13, which is capable of
decoding coded data generated by the image decoding device
( embodiment 13 ) by selectively applying two coding methods used
in the embodiments 1 and 7.
Referring to Fig. 27, coded information and management
information are separated each other at a management-
information separating portion 111 and inputted to a tile
wavelet decoding portion 501 that in turn performs tile-by-tile
decoding of the coded data using the management information
and outputs a decoded image.
The coded data is inputted to a wavelet transform decoding
portion 502 whereby it is wavelet-decoded. The image decoded
by the wavelet transform decoding portion 502 is outputted
directly by a second switch 504 or outputted through a
tile-combining portion 117.
In the wavelet transform decoding portion 502 , the output
of an inverse quantizing portion 114 is applied to a first
inverse wavelet-transform portion 506 through a first switch
503 or to the first inverse wavelet transform portion 506
through a wavelet-coefficient rearranging portion 155.
The operation of the first inverse wavelet-transform
portion 506 is similar to that of the inverse wavelet-
transforming portion 115 in the embodiment 3, so further
description is omitted.
A flag-generating portion 505 extracts flags for
39
controlling the first switch 503 and the second switch 504 from
the management information. With the switches 503 and 504
connected to their terminals 0, the image decoding device
performs the same decoding operation that the embodiment 3 does .
With the switches 503 and 504 connected to the terminals 1,
the image decoding device performs the same decoding operation
that the embodiment 9 does.
The operation of a tile-combining portion 117 is similar
to that of the' same portion of the embodiment 3 described with
reference to Fig. 13, so further description is omitted.
The image decoding device according to the embodiment 13
can process coded image data on a tile-by-tile basis and can
also decode the image by selectively applying two decoding
modes : the simple tile-image-decoding method of the embodiment
3 and the tile-boundary distortionless decoding method of the
embodiment 9, (or the coding method of the embodiment 5,
which can encode each tile without distorting at least upper
and lower or (or/and) left and right of the boundary
thereof).
Figure 28 is a block diagram of another exemplary image
decoding device according to the embodiment 13 of the present
invention, which is capable of decoding image data encoded by
selectively applying two coding methods of the embodiments 1
and 5.
The operation of this device except a tile wavelet
transform decoding portion 601 and a wavelet transform decoding
portion 602 (Fig. 28 ) are similar to that of the device of Fig.
27, so further description of the like portions is omitted.
The wavelet transform decoding portion 602 performs
wavelet decoding of input coded information . The output of an
inverse quantizing portion 114 through a first switch 604 is
inputted to a first inverse wavelet transform portion 506 or
a second inverse transforming portion 603.
The output of the first inverse wavelet transform portion
506 is transferred to a tile composing portion 117 and the
output of the second inverse wavelet transform portion 603 is
transferred to a tile integrating portion 147.
The operation of the second inverse wavelet transforming
portion 603 is similar to the inverse wavelet transform portion
145 of the embodiment 8 described with reference to Fig. 20,
so further description is omitted.
In the tile wavelet decoding portion 601, the wavelet
transform decoding portion 602 performs wavelet decoding of
the coded information and outputs the decoded information to
the tile composing portion 117 or the tile integrating portion
147. A decoded image is now reproduced.
On the other hand, a flag-generating portion 605 extracts
a flag from the management information and controls the
operation of a first switch 604 by the extracted flag. With
the switch 604 connected to its terminal 0, the image decoding
device performs the same decoding operation that the embodiment
3 does. With the switch 604 connected to its terminal 1, the
image decoding device performs the same decoding operation that
the embodiment 8 does.
41
Thus , the image decoding device according to the embodiment
13 can process coded image data on a tile-by-tile basis and
can also decode the image by selectively applying two decoding
modes: the simple image-decoding method of the embodiment 3
and the tile-boundary distortionless decoding method of the
embodiment 8, (or the coding method of the embodiment 5,
which can encode each tile without distorting at least upper
and lower or (or/and) left and right of the boundary
thereof).
Figure 29 is a block diagram of a further exemplary image
decoding device according to the embodiment 13 of the present
invention, which is. capable of decoding image data encoded by
selectively applying three coding methods of the embodiments
1, 5 and 7.
Referring to Fig. 29, this image decoding device differs
from the image decoding device of Fig. 28 by the provision of
a wavelet-coefficient rearranging portion 155 and related
switch circuitry. Since the operation of this device excepting
a tile wavelet transform decoding portion 701 and a wavelet
transform decoding portion 702 (Fig. 29) is similar to that
of the device of Fig. 27, so further description of the like
portions is omitted.
The wavelet transform decoding portion 702 performs
wavelet decoding of input coded information. The output of an
inverse quantizing portion 114 is inputted to a first inverse
wavelet transform decoding portion 506 through a terminal 0
of a first switch or a wavelet coefficient rearranging portion
155 through a terminal 1 of the first switch 703 . Alternatively,
42
it is transferred to a second inverse transforming portion 603
through a terminal 2 of the first switch 703.
The output of the first inverse wavelet-transforming
portion 506 is transferred to a tile-composing portion 117
through a second switch 704 or a decoded image is directly
outputted. The output of the second inverse wavelet-transform
portion 603 is transferred to a tile-integrating portion 147.
The operation of other components is similar to that of like
components of the wavelet-decoding portion 602 (Fig. 28), so
further description is omitted.
In the tile wavelet decoding portion 701, a flag extracting
portion 705 extracts flags for controlling the first switch
703 and the second switch 704 from the management information.
The remaining management information is inputted to the
tile-composing portion 117 and the tile-integrating portion
147.
With the switches 703 and 704 connected to its terminal
0, the image decoding device performs the same decoding
operation that the embodiment 3 does. With the switches 703
and 704 connected to its terminal 1, the image decoding device
performs the same decoding operation that the embodiment 9 does .
When the first switch 703 is connected to its terminal 2, the
device performs, the same decoding operation that the
embodiment 8 does irrespective of the position of the second
switch 704.
Thus, the image decoding device according to the embodiment
43
13 can process coded image data on a tile-by-tile basis and
can also decode the image by selectively applying three
decoding modes: the simple tile-image-decoding method of the
embodiment 3 and the tile-boundary distortionless decoding
method of the embodiments 8 and 9, (or the coding method of
the embodiment 5, which can encode each tile without
distorting at least upper and lower or (or/and) left and
right of the boundary thereof).
An image coding device is described below as an embodiment
14 of the present invention. In this embodiment, tile
management information including information for identifying
(distinguishing) tiles is utilized to realize high-speed
decoding of any objective tile.
Figure 30 is a block diagram showing an exemplary image
coding device according to the embodiment 14. Referring to Fig.
30, a tile wavelet-coding portion 801 performs wavelet encoding
of an input original image on a tile-by-tile basis, and it
generates coded information and management information such
as tile-decomposition information, flag information and
subband information.
An ID generating portion 802 generates ID information for
identifying each tile. Management information generating
portion 803 generates management information by combining the
management information with the ID information. A coded-data
combing portion 804 generates coded data by combining the coded
information with the management information and placing a
tile-information start code at the head of information of each
tile.
44
Figure 31A shows an example of the coded data format that
defines each tile information consisting of a tile-information
start code, management information (tile header) and coded data.
The tile wavelet-coding portion 801 can be commonly used for
image coding devices according to embodiments 1 , 2 , 5 , 6 , 7 ,
10, 12 and 14.
To distinguish tiles into which an original image was
decomposed, ID numbers ( a . g . , 1, 2 , ... ) are as signed to tiles
arranged in a sequence from the top left as shown Fig. 31. Tiles
having ID numbers can be coded in any order and rearranged after
coding. Moreover, the ID generating portion 802 may be omitted
if the order of tiles to be encoded is predetermined.
The location of each tile can be found by its start code
or by its data size (coded information plus tile header).
Figure 32 is a block diagram of another exemplary image
coding device according to the embodiment 14. This image coding
device differs from the device of Fig. 30 only by the provision
of a data-size calculating portion 811, so the portions other
than the calculating portion 811 and the management-
information generating portion 812 are omitted from the scope
of further description.
Referring to Fig. 32, the data-size calculating portion
811 calculates a size of coded data for each tile and outputs
the calculation result. The management information generating
portion 812 prepares management information consisting of
management information, ID information and a coded tile-data
size.
Figure 31B shows an example of a coded information format
in each tile, a coded tile-data size is placed at the head,
and following other, management information (tile header) and
coded information. The coded tile-data size is not necessarily
placed at the top of each tile field. Alternatively, data-
size values for all tiles may be placed together at the top
of the format .
Figure 33 is a block diagram of another exemplary image
coding device according to the embodiment 14. This image coding
device differs from the device of Fig. 32 only by the provision
of a coded-data rearranging portion 821, so the portions other
than the calculating portion 821 and the management-
information generating portion are omitted from the scope of
further description.
Referring to Fig. 33, the coded-data rearranging portion
821 extracts a coded data size for each tile from coded data
prepared by a code-data combing portion 804 and puts the
data-size value at the head of coded data and arranges other
remaining data in a given order, then outputs a sequence of
the coded data.
In case of an exemplary coded-data format shown in Fig.
31C, the location of any objective tile can be easily determined
by summing data-size values from the top tile to a just
preceding one.
Figure 34 shows another exemplary image coding device
46
according the embodiment 14, which can realize the same affects
as the above and which differs from the above device of Fig.
32 by the provision of a coded data storaging portion 831 and
a management information storaging portion 832 . Therefore, all
components other than the coded-data storing portion 831, the
management information storaging portion 832 and a coded data
combining portion 833 are omitted from further description.
Referring to Fig. 34, coded information outputted from a
tile wavelet coding portion 801 are temporally stored in the
coded-data storaging portion 831. The management information
storaging portion 832 stores tile management information
generated by the management-information generating portion
812. It extracts tile-size data from the tile-management-
information, sends the data to the coded data combining portion
833 and then outputs the remaining management information.
First, the coded data-combining portion 833 outputs all
the data size values for all tiles and then outputs remaining
management information combined with coded information.
According to the embodiment 14 of the present invention,
it is possible to immediately retrieve coded information of
any desirable tile to be decoded.
An image decoding device is described below as another
embodiment 15 of the present invention. Figure 35 is a block
diagram of an image decoding device according to the embodiment
15 , which is capable of decoding coded data encoded and supplied
by the image coding device according to the embodiment 14.
47
Referring to Fig. 35, an objective-tile deciding portion
903 decides an ID of an objective tile to be decoded according
to the user's request. A management-information separating
portion 906 retrieves a start code indicating the head of the
ob jective-tile coded information in a coded-data sequence and
separates the ob jective tile management information from the
objective-tile coded-information.
Based on the management information, a data skip control
portion 902 decides whether the ID of the tile to be decoded
matches the ID. decided by the deciding portion 903. If two IDs
match, the portion 902 turns on both a first switch 905 and
a second switch 904. Consequently, a tile wavelet-decoding
portion 901 can decode the selected tile only.
If the tile management information includes each tile-
data size, the management-information separating portion 906
has no need to search tile-data heads and can find the location
of the objective tile data head by skipping the unnecessary
amount . The tile wavelet-decoding portion 901 can be commonly
used in the image decoding devices according to the embodiments
3, 4, 8, 9, 11, 13 and 15.
According to the embodiment 15, it is possible to
immediately retrieve and decode the coded data of the object
tile by using only the tile head management information without
decoding any other coded data.
An image coding device is now described below as another
embodiment 16 of the present invention, which can provide coded
48
tile images that can be decoded immediately to reproduce a
single objective tile image as well as adjacent tile images
by using tile-management information including neighbors'
information.
Figure 36A is a block diagram of an exemplary image coding
device according to the embodiment 16 . This device differs from
the embodiment 14 of Fig. 30 by the addition of an adjacent
tile ID deciding portion 841 and by the operation of a
management-information generating portion 841. Other portions
are similar to. those of the embodiment 14 and omitted from the
scope of further description.
A tile wavelet-coding portion 841 can be commonly used in
the image coding devices of the embodiments 5, 6, 7, 10, 12
and 14.
Referring to Fig. 36A, the adjacent tile ID deciding
portion 841 decides IDs of adjacent tiles necessary for
decoding an objective tile according to tile decomposition
information, f lag information, subband information and tile
IDs produced by an ID generating portion 802. A
management-information generating portion 842 prepares
management information containing tile decomposition
information, f lag information, subband information and tile
ID with adjacent tiles IDs.
Since all adjacent tiles necessary for coding an objective
tile are not necessarily given IDs, the number of adjacent IDs
to be produced by the peripheral tile ID deciding portion 841
49
may be limited to, for example, two neighbors existing left
above and left below the objective tile as shown in Fig. 36B.
In the coded data format of Fig. 31A, the management
information (tile header) may include an objective tile ID and
adjacent tiles' IDs.
Figure 37 is a block diagram of an exemplary image coding
device according to the embodiment 16 of the present invention,
which is intended to encode tile images that-may be rapidly
retrieved for decoding by using management information
including each objective tile ID with neighbors' IDs. This
image coding device is similar in construction to embodiment
14 of Fig. 34 but differs from the embodiment 14 by the absence
of the management information storing portion 832 and by the
presence of a data-size storing portion 853, relative position
calculating portion 852 and information storing buffer 854.
The operation of this image coding device is similar to
the embodiment 14 except the operation with the data-size
storing portion 851,relative position calculating portion852,
information storaging buffer 854, management information
generating portion 853 and ID generating portion 855 . Therefore,
like components are not described further.
Referring to Fig. 37 , coded information outputted from a
tile wavelet coding portion 801 are all stored in a coded data
storaging buffer 831. Tile-decomposition information, flag
information and subband information from the tile wavelet
coding portion 801 are all stored in the information storaging
buffer 854. Tile-data-size values outputted from the data-
size calculating portion 811 are all stored in the data-size
storing portion 851.
The ID generating portion 855 outputs ID information to
identify each tile and controls the information storing buffer
854, the data-size storing portion 851 and the coded' data
storaging buffer 831 to output information on a tile-by-tile
basis. The data-size storing portion 851 receives a tile ID
and outputs a data-size value of the tile specified by the
received ID to the management information generating portion
853. It also provides the relative-position calculating
portion 852 with the tile-data-size necessary for calculating
positions of neighbors relative to the tile having the ID.
The relative-position calculating portion 852 calculates
the positions of coded information of the adjacent tiles
relative to an objective tile by using the data-sizes of the
input tiles and outputs the calculation results . The management
information generating portion 853' generates management
information by using input information such as tile ID
information, tile-decomposition information, flag
information, subband information, tile-data-size values,
relative positions of adjacent tiles, etc. It outputs the
prepared management information to the coded data combining
portion 833.
The above system can produce coded data of tile images,
which can be effectively decoded at a high speed without
51
decoding all coded data in such a way that coded data of an
objective tile and necessary adjacent tiles may be retrieved
and decoded by decoding only management information located
at the head of the decodable coded data.
An image decoding device is described below as another
embodiment 19 of the present invention. Figure 38 is a block
diagram of the image decoding device according to
embodiment 19 , which is used for decoding coded data produced
by the image coding device according to the embodiment 18. This
device is similar to and differs from the embodiment 15 of Fig.
35 only by the addition of a buffer 911. All components other
than the buffer 911 and a skip-and-read control portion 912
operate similarly to like components of embodiment
15. So, they are omitted from the scope of further description.
Referring to Fig. 38, input coded data is temporally stored
in the buffer 911 wherefrom it is subsequently outputted later.
The skip-and-read control portion 912 extracts an ID for an
objective tile according to the input management information.
When the extracted ID matches the ID of the objective tile or
the ID of a related adjacent tile, this control portion 912
causes a first switch 905 and a second switch 904 to turn ON.
If the management information contains IDs of adjacent
tiles necessary for decoding the ob jective tile, the control
portion 912 controls the buffer storage 911 to output the coded
data of the adjacent tiles. Consequently, the tile
wavelet-decoding portion 901 can decode a specified tile and
52
necessary neighbors.
If the predetermined number ( a . g . , two dotted tiles in Fig .
36B) of peripheral IDs decoded in the management information
is smaller than the number of necessary peripheral tiles ( a . g . ,
six unshaded neighbors in Fig. 36B), IDs of the remaining
necessary neighbors are decided from the decoded IDs of the
adjacent tiles.
The tile wavelet-decoding portion 901 can be commonly used
in embodiments 8, 9, 11, 13 and 15.
The above system can immediately decode any objective tile
and necessary adjacent tiles by decoding only the management
information put at the head of the coded data. It has no need
of decoding all the coded data.
THE POSSIBLE INDUSTRIAL APPLICATIONS
As described herein, following aspects are brought
according to the present invention.
In one aspect of the present invention, an image coding
device can independently encode each of the tiles of an original
image, thus providing coded tile images that can be separately
treated thereafter. If any coded tile must be further
processed, it can be separately processed and encoded again
with no need of using adjacent pixels. Thus, simple independent
encoding and decoding of image tiles is realized.
In another aspect of the present invention, an image
decoding device can decode only a desirable coded tile image
53
with no need of decoding any other coded data, thereby
minimizing the processing load.
In another aspect of the present invention, in spite of
increasing of the coded-data size due to encoding an objective
tile image including adjacent pixels, an image decoding device
decodes the coded tile image by superposing adjacent pixel
values on overlaps, suppressing possible boundary distortion
of the tile image .
In still another aspect of the present invention, an image
coding device can encode tile images using pixel information
on neighboring tiles, achieving high efficiency of image
encoding using the correlation between tiles. This can also
suppress possible boundary distortion of the tile images.
In another aspect of the present invention, an image coding
device can effectively encode a part (plural tiles) of a whole
image by performing wavelet transform of only selected tiles,
and its wavelet transform is very compact.
An image decoding device responding to the above can also
realize compact inverse wavelet transform of coded tile images .
In a further aspect of the present invention, an image
coding device can decide exclusion of distant pixels from the
scope of adjacent pixels for calculation. This minimizes the
number of filtering operations and wavelet-transform
operations.
A whole image is wavelet transformed at a time and then
wavelet transformed coefficients are rearranged to compose
54
respective tiles. This eliminates the need of iterating the
wavelet-transform for each tile.
In another aspect of the present invention, an image
decoding device can rearrange coded data ( decomposed for each
tile) corresponding to an objective tile and then perform
inverse wavelet transform of the coded data at a time, thus
eliminating the need of repeating inverse wavelet transform
for each tile.
Conventional arts demand a large capacity of a memory for
holding wavelet transformed coefficients to correspond to
resolution of an original image. In contrast to the above, an
image coding device according to one aspect of the present
invention can use, irrespective of the original image size,
a memory which can store only wavelet transform coefficients
for capacity corresponds to the size of a tile or tiles for
a tile or tiles being currently encoded. This can realize a
considerable saving of memory capacity needed.
In another aspect of the present invention, an image
decoding device can also use a memory having the capacity
limited to a tile size for storing wavelet transform
coefficients .
In a further aspect of the present invention, an image
coding device can conduct wavelet transform by selectively
applying plural suitable filters for decomposing an objective
tile image into subbands, thus realizing optimal wavelet
transform of the objective tile with the best balance between
the image quality and the processing load.
In a further aspect of the present invention, an image
decoding device can conduct inverse wavelet transform by
selectively applying plural suitable filters for composing
subbands, which respond to the subband decomposing filters.
Thus, the device realizes the optimal inverse wavelet transform
of each tile to be decoded.
In a still further aspect of the present invention, an image
decoding device can selectively apply two wavelet transform
modes : one for each tile with or without adjacent pixels and
the other for each image with or without adjacent pixels . Two
modes can be selected in accordance with the input image quality
to minimize the increase of transform operations and
deterioration of the image quality.
In another aspect of the present invention, an image
decoding device can easily retrieve coded information of an
objective tile to be decoded by searching management
information well organized for identifying every tile.
Therefore, it can realize high-speed retrieval and effective
decoding of only plural tiles necessary for reproducing an area
(several tiles) of a whole image.
In still another aspect of the present invention, an image
decoding device can easily retrieve coded information of an
objective tile and coded information of adjacent tiles
necessary for decoding the objective tile by using management
information well organized for identifying every tile.
56
Therefore, only an area (several tiles) of a whole image
can be rapidly reproduced by decoding only necessary tiles.
High efficient coding/decoding of image tiles using
adjacent pixel information is also achieved by utilizing the
effect of correlative effect of the pixels.
57
