Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02436437 2003-07-28
DESCRIPTION
METHOD AND SYSTEM FOR COMPRESSING MOTION IMAGE
INFORMATION
Technical Field
The present invention relates to a method and system for compressing
motion image information, which can compress, with a high compression ratio
and at a high speed, data or image information that can be subjected to
predictive coding, and which can improve image quality.
Background Art
Conventionally, it is a general procedure to convert an image signal to
another type of a signal, and assign suitable codes to that converted signal
based on statistic characteristics of the converted signal, and transmit the
resulting coded signal. In this case, so-called predictive coding, which may
compress information with a large compression ratio, is performed for a
redundant image within a frame or an image including a regular pattern or a
plain pattern, in such a manner that: since there may be a high correlation
between adjacent pixels, it is possible to predict, to a certain degree, the
next
pixel value to be coded from a pixel value which has already been coded only
the components that could not be predicted are extracted and encoded.
In the case of a motion image in a videophone or the like, since adjacent
frames of images are often very similar to each other, the temporal changes
are limited accordingly, such temporal redundancy may be removed by
CA 02436437 2003-07-28
2
inter-frame predictive coding, which performs prediction between frames. .At
this time, so-called block-based coding may be generally performed in such a
manner that: a block code is employed in which one codeword is assigned to a
single symbol each frame is divided into a plurality of pixel blocks utilizing
S the characteristic that the luminance difference within each block is
smaller,
information is compressed.
Huffman coding is known as a method for generating a high efficiency
code, which is one of entropy coding that may achieve data compression by
assigning a high efficiency code to the converted signal. A representative
thereof is arithmetic coding, which generates a codeword after another through
an arithmetic calculation by dividing a probabilistic number line into
segments
in accordance with the occurrence probability of each sequence of symbols, and
determining to accept the binary decimal indicating a location in a segment as
a code for the sequence of symbols.
A conventional three-step block coding system for efficiently coding an
image signal comprises the steps of sampling, transforming, and quantization.
In order to retain two-dimensional resolution and high-frequency components
for a given image signal, it is generally required to perform sampling at a
frequency twice the highest frequency component.
With MPEG, it is preferable that the coding efficiency be as high as
possible so that images with a high amount of information can be compressed.
Accordingly, there are the conventional forward-predictive coding (P frame
based processing), which uses as a predictive signal only the past image
signal
that has been already encoded, and the bi-directional predictive coding (B
CA 02436437 2003-07-28
3
frame based processing), which uses as a predictive signal a future image
signal as well as the past image signal. The conventional inter-frame
predictive coding performs transmission of the difference signal between an
input image signal and corresponding predictive image signal, and the
decoding side performs addition of the transmitted difference signal to the
already-decoded predictive image signal so as to reconstruct the original
image.
In this manner, it is impossible for the decoding side to reconstruct on the
inter-frame basis if the predictive image signal is not provided. Accordingly,
not using the past and future image signals as predictive signals, but using
an
intra-frame coded I frame (i.e., a reference frame that allows for
reconstruction
of an image from solely that frame), this I frame is inserted into a sequence
of
frames at a fixed interval, enabling reconstruction of an image partially
through the sequence thereof and solving possible data errors.
However, since the conventional image signal compression technique
employs the complicated block-based coding procedure, it is difficult to
compress with a high compression ratio and at a high speed, image data such
as audio information that can be subjected to predictive coding. When
difference information is generally compressed through a motion image
compression procedure, that is, when successive values A1 and A2 are
expected to be similar to each other and when the value A1 is known before the
value A2 occurs, assuming that the probability of occurrence of the difference
between A2 and A1 being equal to or near zero be high, compression is
performed using the conventional Huffman coding or the arithmetic coding
accordingly, if A1 and A2 can each take one of values 0,..., n, the difference
CA 02436437 2003-07-28
4
between A2 and A1 can have one of 2n + 1 values, thus 2n + 1 Huffman
codewords are necessary. Since there are actually n possible values for A2,
but all of the n codes are not locally used, redundant codewords may be
generated. Moreover, there is a problem that when the difference between
frames is large, image quality intensely deteriorates and it is impossible to
provide high image quality.
In addition, if a larger block size is used, compression ratio may be
enhanced however the detail of the original image may be lost causing image
quality to deteriorate. A phenomenon wherein fine lines may be completely
lost occurs when the original image is configured by such fine lines, the
intensity thereof differing from that of a fixed colored background.
Furthermore, since the I frame to be periodically inserted into a
sequence of frames is subjected to intra-frame coding, the coding efficiency
is
worse as compared to that of inter-frame coding, which encodes the difference
between frames, so that generated amount of information increases therefore,
in such a case of a high speed communication line not being available, the
frequency of insertion of I frames is limited. Moreover, since the amount of
data within an I frame is between double to ten times that of data within a
differential frame, that technique is against a fixed rate required for
enabling
communication. That is, conventionally, since I frames are periodically
inserted into a sequence of frames, the processing time is fairly long so that
displaying of the reconstructed image is delayed very long. In addition, since
the amount of data itself is large, probability is high that data error
impossible
to be restored occurs within an I frame. Moreover, in case where such an error
CA 02436437 2003-07-28
occurs causing reconstruction (or decoding) of an I frame to be impossible,
the
reconstruction processing halts until the subsequent I frame is reached if a
dedicated means to solve the problem is not provided. For example, in case
where data error occurs due to a cause, its initially small adverse influence
5 may be amplified over the entire many frames in the worst case, the
reconstruction processing halts. Furthermore, with the conventional
technique for inserting I frames at a fixed interval, it is needed, when the
reconstruction processing starts at the frame at a certain temporal position,
to
search for the nearest I frame by some means, reconstruct the corresponding
image therefrom, and display the reconstructed image after the frame at the
target temporal position is reached however, this searching for the I frame
takes much time. If a certain dedicated means to solve the aforementioned
problems is provided, the corresponding burden to be imposed on the decoding
process naturally increases. In addition, since a heavy burden will be imposed
on the process of reconstructing I frames, an additional function capable of
processing the I frames is required for the process.
The present invention is provided considering the above problems, and
its first object is to provide a method and system for compressing motion
image
information, which can compress with a high compression ratio and at a high
speed, data or image information that can be subjected to the predictive
coding,
and which can improve image quality.
The second object of the present invention is to provide a method and
system for compressing motion image information with the detail of the
original image being preserved and without deterioration of image quality
CA 02436437 2003-07-28
6
even if the compression ratio is enhanced by enlarging the block size.
Further, the third object is to provide a method and system for
compressing motion image information, which can easily display the
reconstructed image at an arbitrary temporal position by preventing an
initially adverse influence due to an occurrence of data error during a
reconstruction process from prevailing over all of the many frames and in turn
the reconstruction process from halting, without much time being taken for
first searching for the nearest I frame by some means when the reconstruction
process starts at the frame at an arbitrary temporal position and then the
corresponding image is reconstructed therefrom.
DISCLOSURE OF INVENTION
According to a motion image information compression method of a first
embodiment of the present invention, which compares spatially adjacent pixels
within a frame to each other or compares temporally adjacent pixels between
frames, outputs the resulting difference information for pixels, stores in a
bit
map, information regarding on whether or not the output difference
information is greater than a given parameter (threshold), and compresses the
information stored in the bit map that is greater than the parameter
(threshold), thereby reducing redundant information by dividing an image
within a frame into blocks and approximating (substituting) each divided block
by a plane represented by three components for pixels within each block, the
aforementioned problems are solved.
On the other hand, a system of compressing motion image information,
CA 02436437 2003-07-28
7
according to the first embodiment of the present invention, comprises: bit map
information recording means for comparing spatially adjacent pixels within a
frame to each other or comparing temporally adjacent pixels between frames
outputting the resulting difference information for pixels, and storing in a
bit
map, information regarding on whether or not the output difference
information is greater than a given parameter (threshold), and information
compression means for compressing the information stored by the bit map
information recording means that is greater than the parameter (threshold),
thereby reducing redundant information the system further comprises block
approximation means for dividing an image within a frame into blocks before
an inter-frame compression is performed, and approximating (substituting)
each divided block by a plane represented by three components for pixels
within each block, thereby solving the aforementioned problems.
According to a motion image compression method of the second
embodiment of the present invention, the intra-frame compression is
performed by compressing the entire image in an n x m pixels block unit (n
and m are integers, respectively) using a intra-frame compression method,
comparing pixels between the original image and the image expanded after
compressed outputting the resulting difference information of each pixel, and
if
a pixel that caused a larger difference than a given parameter (threshold) to
occur exists, repeatedly using a smaller block size for a portion or a
surrounding area including this pixel is performed until a designated
minimum block size is reached thereby solving aforementioned problems.
Furthermore, according to a motion image compression system of the
CA 02436437 2003-07-28
g
second embodiment of the present invention, block approximation means
performs an intra-frame compression by compressing the entire image in an n
x m pixels block unit (n and m are integers, respectively) using a intra-frame
compression method, comparing pixels between the original image and the
image expanded after compressed, outputting the resulting difference
information of each pixel, and if a pixel that caused a larger difference than
a
given parameter (threshold) to occur exists, repeatedly using a smaller block
size for a portion or a surrounding area including this pixel until a
designated
minimum block size is reached thereby also solving the aforementioned
problems.
According to a motion image compression method of the third
embodiment of the present invention, an intra-frame coded I frame (i.e., a
reference frame solely from which an image can be reconstructed) is used the I
frame is spatially divided into I blocks and no I block is inserted in any
block
within the frame that has been updated due to difference between frames
being greater than a given parameter (threshold) within a specific period of
time when dispersing the I blocks between each frame along the temporal axis
thereby also solving the aforementioned problems.
Furthermore, a motion image compression system of the third
embodiment of the present invention comprises I block insertion means, which
uses an intra-frame coded I frame (i.e., a reference frame solely from which
an
image can be reconstructed), spatially divides the I frame into I blocks, and
disperses the I blocks between each frame along the temporal axis the I block
insertion means does not insert any I block in a block within the frame which
CA 02436437 2003-07-28
9
has been updated due to difference between frames being greater than a given
parameter (threshold) within a specific period of time thereby also solving
the
aforementioned problems.
According to the motion image information compression method and
system thereof, since a block conversion procedure is omitted, compression
with a high compression ratio and at a high speed, of data or image
information that can be subjected to predictive coding can be performed
thereby improving image quality. With the conventional technique, in
particular, when difference between frames is large, image quality drastically
deteriorates however, according to the present invention deterioration of that
image quality can be reduced. More specifically, according to the first
embodiment of the present invention, it is possible to provide linear change
in
image quality without drastic deterioration of image quality due to a
threshold
for a block. Accordingly, adjustment of communication bit rate can be easily
performed without deterioration of image quality, and in addition,
improvement by approximately -20 % to -50 % of the compression ratio can be
made with image quality maintained as is. Moreover, the adaptive Huffman
coding and the adaptive arithmetic coding collectively perform the predictive
coding procedure including the conventional difference information generation
and the Huffman coding and/or the difference information generation and the
arithmetic coding thereby generating efficient codewords and efficiently
compressing data such as image information that can be subjected to
predictive coding. Furthermore, the reduced (compressed) data according to
the first embodiment of the present invention is used to define a plane, and
CA 02436437 2003-07-28
when it is expanded, it represents a plane with a gradation.
According to an motion image information compression method and
system of the second embodiment of the invention, even in the cases where the
compression ratio is improved using a larger block size, the detail of the
5 original image is not lost, and accordingly deterioration of image quality
can be
reduced. Even in cases of an original image configured by greatly deferent
intensities of fine lines with a fixed colored background, it is possible to
prevent that fine lines from being completely lost.
According to a motion image information compression method and
10 system thereof of the third embodiment of the present invention, when an I
frame is spatially pre-divided into blocks, and when the divided I blocks are
dispersed between each frame along the temporal axis, since no I block is
inserted in any block within the frame that has been updated due to difference
between frames being greater than the parameter (threshold), it is possible
for
the image reconstruction processing to begin reconstruction since a
predetermined number of previous frames from which an image can be
completely reconstructed, and display a reconstructed image after the frame in
the target temporal position is reached thus without time-consuming search
for an I frame, a reconstructed image can be easily displayed in an arbitrary
position. In addition, since the amount of the distributed data in a
communication server and/or data transmission path is temporally
uniformized during motion image delivery, a higher transmission performance
for content delivery than that with the conventional technique can be
obtained.
Moreover, since on the reception and reconstruction side change in the
CA 02436437 2003-07-28
11
received amount per unit time is small, a necessary amount of buffering
memory can be reduced, expected reconstruction loads are regulated, and even
a system with low capacity can stably reconstruct. Moreover, since possible
influence of data errors upon reconstruction is small, it is possible to
continue
to reconstruct with data error neglected thereby not requiring the delivery
side system to re-send data, and reducing the burden on the delivery side.
Moreover, it is also possible to easily provide multicasting distribution
capability, etc. for the motion image broadcasting.
BRIEF DESCRIPTION OF DRAWINGS
Fig. 1 is a block diagram showing the outline of a structure for compressing
motion image information.
Fig. 2 is a block diagram detailing the structure for compressing motion image
information in Fig. 1.
Fig. 3 illustrates an example of a specific structure for coding.
Fig. 4 is an explanatory block diagram showing an example of a specific
structure for decoding.
Fig. 5 is an explanatory figure showing a plane represented by three pieces of
data: intensity Z of a pixel within a block, the gradient of the block in the
X
direction, and the gradient of the same block in the Y direction, which are
used
to approximate the corresponding divided image block.
Fig. 6 is a plan view of an image explaining the operation of using a smaller
block size for a portion or a surrounding region of a pixel that causes larger
difference than a given parameter (threshold) to occur.
CA 02436437 2003-07-28
12
Fig. 7A and 7B show an image explaining an inter-frame compression
procedure, wherein Fig. 7A is a plan view of frame t and Fig. 7B is a plan
view
of frame t + 1
Fig. 8 is a plan view showing a plurality of I blocks that configures an I
frame
Fig. 9 is an explanatory figure showing a state where I blocks are inserted
between frames
Fig. 10 is a flowchart showing the step of performing an intra-frame
compression
Fig. 11 is a flowchart showing the step of performing an inter-frame
compression and
Fig. 12 is a flowchart showing the step of performing I block insertion.
DESCRIPTION OF THE REFERENCE NUMERALS
P .. . parameter (threshold)
1 .. . analog to digital converter
2 ... buffer
3 . .. encoder/compression unit
4 ... bit map information recording means
5 ... information compression means
6 .. . entropy coding means
7 ... I block insertion means
8 ... I block generation means
12 ... comparison means
13 ... current frame data
CA 02436437 2003-07-28
13
14 ... previous frame data
BEST MODE FOR CARRYING OUT THE INVENTION
To begin with, a method of compressing motion image information
according to a first embodiment of the present invention is described.
The present invention involves a motion image information compression
method for comparing spatially adjacent pixels within a frame to each other or
comparing temporally adjacent pixels between frames outputting the resulting
difference information for pixels, storing in a bit map, information regarding
on whether or not the output difference information is greater than a given
parameter (threshold), and compressing information stored in said bit map
that is greater than said parameter (threshold), thereby reducing redundant
information wherein am image within a frame is divided into blocks, and each
divided block is approximated (substituted) with a single plane represented by
three components for pixels within said block before the inter-frame
compression processing starts.
Further, information stored in the bit map not greater than parameter
(threshold) P is processed (deleted) as a changeless pixel.
Furthermore, according to a block approximation method of configuring
a single plane represented by three components for pixels, the average and the
method of least square are utilized.
Further, according to the intra-frame compression procedure, said plain
is represented by three pieces of data: the intensity of a pixel within a
block,
the gradient of intensities within the block in the X direction, and the
gradient
CA 02436437 2003-07-28
14
of intensities within the block in the Y direction.
The information stored in the bit map is compressed by at least one
binary image coding method selected from the group consisting of run length
coding, modified READ (MR, MMR) coding, modified Huffman (MH) coding,
and JBIG coding.
Information greater than parameter P (threshold) is compressed using
the adaptive Huffman coding, which utilizes as many Huffman tables as the
expected amount of information.
Redundant information between frames is further reduced using an
entropy coding.
The entropy coding is performed either through the adaptive Huffman
coding procedure, which encodes utilizing a table selected from as many
Huffman tables as an expected amount of information, or the adaptive
arithmetic coding procedure, which encodes utilizing a table selected from as
many arithmetic tables as an expected amount of information.
This coding procedure is performed based on difference information
between pixels.
This difference information is the difference output through comparison
of pixel t and pixel t - 1 between frames.
Furthermore, difference information output through comparison of a
block of n x m pixels (where n and m are integers of 2 or more) and
corresponding block of the same between frames is utilized.
Furthermore, difference information output through comparison of pixel
t and pixel t - 1 between said frames is utilized, where a block is configured
by
CA 02436437 2003-07-28
n x m pixels (n and m are integers of 2 or more) within a frame.
Furthermore, with n x m pixels between frames, n denotes 2K (K is a
whole number), and m denotes 2K' (K' is a whole number.)
Furthermore, intra-frame compression may be performed while
5 changing the size of blocks divided within the same frame, before the
inter-frame compression procedure starts.
Next, a system for compressing motion image information according to a
first embodiment of the present invention is described.
The present invention involves a motion image information compression
10 system comprising: bit map information recording means 4 for comparing
spatially adjacent pixels within a frame or comparing temporally adjacent
pixels between frames, outputting the resulting difference information of the
pixels, and storing in a bit map, information regarding on whether or not the
output difference information is greater than a given parameter (threshold) P,
15 and information compression means 5 for compressing the difference
information stored by said bit map information recording means 5 that is
greater than said parameter (threshold) P, thereby reducing redundant
information said system further comprises a block approximation means for
dividing an image within a frame into blocks, and approximating
(substituting) each divided block by a single plane represented by three
components for pixels within said block.
Furthermore, information compression means 5 processes (deletes) the
information stored by bit map information recording means 4 not larger than
parameter (threshold) P as a changeless image.
CA 02436437 2003-07-28
16
Furthermore, the block approximation means approximates a single
plane represented by three components for pixels using the average and the
method of least square.
In addition, with the block approximation means, the plain is
represented by three pieces of data: the intensity of a pixel within a block,
the
gradient of the intensities of the block in the X direction, and the gradient
of
the intensities of the block in the Y direction.
Further, the information stored by bit map information recording means
4 is compressed by at least one binary image coding method selected from the
group consisting of run length coding, modified READ (MR, MMR) coding,
modified Huffman (MH) coding, and JBIG coding.
Furthermore, information compression means 5 for compressing the
information that is greater than parameter (threshold) P performs an adaptive
Huffman coding using as many number of Huffman tables as a predictive
amount of information.
Entropy coding means 6, which may reduce redundant information
between frames, is further provided this entropy coding means 6 performs
either the adaptive Huffman coding, which encodes utilizing a table selected
from as many Huffman tables as an expected amount of information, or the
adaptive arithmetic coding, which encodes utilizing a table selected from as
many arithmetic tables as an expected amount of information.
Furthermore, difference information stored in bit map information
recording means 4 is a difference output through comparison of pixel t and
pixel t - 1 between the frames, where a block is configured by n x m pixels (n
CA 02436437 2003-07-28
17
and m are integers of 2 or more) within a frame.
In the following, an embodiment of a method and system for
compressing motion image information, according to the first aspect of the
present invention, is described.
Fig. 1 is a block diagram showing the outline of a structure for
compressing motion image information. A composite analog signal output
from a device such as a video camera, a disk player, or a video cassette
player
according to the NTSC standard, is converted to a digital signal to represent
a
single line of a video frame by an analog-to-digital converter 1, and
digitally
output to buffer 2, where it is in turn stored. Note that although it is
disclosed that the analog signal output from the NTSC device is converted to a
digital signal by the analog-to-digital converter 1 and the resultant digital
signal is output and stored in the buffer 2, the present invention is not
limited
to this structure. In other words, according to the present invention, any
video
signal including common video signals output from any of various types of
devices can be efficiently compressed.
As shown in Fig. 1, bit map information storage circuit 4, which
sequentially compares pixel t and pixel t - 1 between frames, and stores in a
single bit based bit map, information regarding on whether or not the
resulting
difference is greater than parameter (threshold) P, is provided. The
comparison of this pixel t and pixel t - 1 is performed based on a pixel
component (i.e., intensity or hue). This is to temporally compare a pixel
(pixel
t) in the current frame and corresponding pixel (pixel t - 1) in the previous
frame where t denotes time. Accordingly, the difference between pixels t and
CA 02436437 2003-07-28
Ig
t - 1 stored by bit map information recording means 4, which is greater than
parameter (threshold) P, is compressed, but others are determined to be of
changeless pixels and then processed (deleted). That information (difference)
greater than parameter P (threshold) is compressed by information
compression means 5 using the adaptive Huffman coding, which utilizes as
many Huffman tables as the expected amount of information. Entropy coding
means 6, which performs comparison of spatially or temporally adjacent pixels
outputting the resulting difference information, and which then performs the
adaptive arithmetic coding utilizing a arithmetic table, which is selected
from
as many arithmetic tables as, for example, an expected amount of information
based upon predictive information so that redundant information between
frames can be reduced, is provided.After encoding is performed by compression
encoder 3, a block of data within each frame is then transmitted to a memory
of bit map information recording means 4, as shown in Fig. 2. The current
1 S frame data 13 and the previous frame data 14 that is delayed by a single
frame
time are then stored. Afterwards, comparator 12 determines the current frame
data 13 and the previous frame data 12 in terms of redundancy between
frames and calculates the difference thereof. That is, each encoded block is
compared to the corresponding block in the previous frame. Each block is
marked with a single bit identifying whether or not that each block is changed
from the corresponding previous block. Through this procedure, a frame bit
map with a single bit per block is generated. Herein, the bit map for each
frame is distinguished from another bit map by performing comparison
between frames.
CA 02436437 2003-07-28
19
This embodiment uses as a basic technique, the intra-frame compression
technique in which the size of blocks is not changed. As shown in Fig. 5, an
image within a frame is pre-divided into blocks, and every pre-divided block
is
approximated (substituted) with a single plane represented by three pieces of
data: the intensity Z of a pixel in each block, the gradient of the
intensities
within each block in the X direction, and the gradient of the intensities in
each
block in the Y direction. More specifically, with the intra-frame compression
procedure, an image is first divided into a plurality of blocks, and each
block is
substituted with a single plane that approximates that each block. This plane
may be represented by three components for pixels in each block such as
intensity z, gradient x of the intensities in the X direction, and gradient y
of
the intensities in the Y direction. Alternatively, that plane may be
represented
by intensity z of a pixel within a block, the gradient of the intensities of
pixels
between blocks in the X direction, and the gradient of the intensities of
pixels
between blocks in the Y direction. The average and the method of least square,
for example, may be used for approximation. The resulting decreased
(compressed) data represents a plane by expanding that data, the plane with
gradation is obtained. When a block is configured by s pixels, an expected
compression ratio within a single frame is 3/ s~ the compression ratio
increases
while s increases, however, image quality deteriorates. It is noted that the
size
and shape of a block is of n x m pixels where n and m are any one of integers.
Furthermore, with n x m pixels between frames, n may denote 2K (K is a whole
number), and m may denote 2K' (K' is a whole number.)
Next, the basic technique used for inter-frame compression according to
CA 02436437 2003-07-28
this embodiment is described.
More specifically, according to the first method for inter-frame
compression, the block in frame t - 1 positioned at the same location as that
in
frame t is intra-frame compressed, and z(t + 1), x(t + 1), and y(t + 1) are
S obtained in terms of three components: z denoting the intensity of a pixel,
x
denoting the gradient of the intensities in a block in the x direction, and, y
denoting the gradient of the intensities in the block in the y direction. The
sum-square-mean error is calculated between a group of z(t), x(t), and y(t)
and
a group of z(t + 1), x(t + 1), and y(t + 1), and then compared with threshold
P.
10 As a result, if it exceeds threshold P, determination such as 'THERE IS
DIFFERENCE' is made. Alternatively, each of a group of z(t), x(t), and y(t),
and a group of z(t + 1), x(t + 1), and y(t + 1) are compared to a group of
thresholds Pz, Px, and Py~ If the resulting difference exceeds threshold P,
determination of 'THERE IS DIFFERENCE' is made. If the determination of
15 'THERE IS DIFERENCE' is made, the portion in the bit map corresponding to
the block in the frame is marked.
In the former case, a single bit map is used, whereas three bit maps are
used in the latter.
This bit map comprises an array including 0 and/or 1 (i.e., binary data),
20 and it is compressed using, for example, run-length coding. In addition,
the
pieces of difference data Oz(t) = z(t + 1) - z(t), Ox(t) = x(t + 1) - x(t),
and ~y(t) _
y(t + 1) - y(t) are entropy compressed. It is noted that according to the
first
method, since expansion is not performed, the burden imposed on calculation
is light, however, calculation errors may be accumulated.
CA 02436437 2003-07-28
21
According to the second method for inter-frame compression, data
compressed using a basic technique for the above-mentioned inter-frame
compression is expanded, and pieces of pixel data configuring a block are
reconstructed. A sum-square-mean error between the respective pieces of pixel
data each located at the same place within the same block in the next frame t
+ 1 and corresponding reconstructed pieces of pixel data is calculated and
compared to threshold P. As a result, if it exceeds threshold P, determination
of 'THERE IS DIFFERENCE' is made. If the determination of 'THERE IS
DIFERENCE' is made, the portion in the bit map corresponding to the block in
the frame is marked. This bit map comprises an array including 0 and/or 1
(i.e., binary data), and it is compressed using, for example, run-length
coding.
In addition, the pieces of difference data Oz(t) = z(t + 1) - z(t), Ox(t) =
x(t + 1) -
x(t), and Dy(t) = y(t + 1) - y(t) are entropy compressed. It is noted that
according to the second method, since expansion is performed, the burden
imposed on calculation is heavy, however, calculation errors are not
accumulated.
According to the third method for inter-frame compression, a
sum-square-mean error between the respective pieces of pixel data within a
block in the current frame t and corresponding pieces of pixel data each
located
at the corresponding identical place within the corresponding identical block
in
the next frame t + 1 is calculated and compared to threshold P. As a result,
if it
exceeds threshold P, determination of 'THERE IS DIFFERENCE' is made. If
the determination of 'THERE IS DIFERENCE' is made, difference OP from the
corresponding pixel data located at the corresponding identical place within
CA 02436437 2003-07-28
22
the corresponding identical block in the next frame t + 1 is calculated and
inter-frame compressed. The portion in the bit map corresponding to the
block in the frame is marked. This bit map comprises an array including 0
and/or 1 (i.e., binary data), and it is compressed using, for example, run-
length
coding. Difference data DP is entropy compressed. It is noted that according
to the third method, since compression is performed after determination of
difference is made, the amount of the calculation is the least and calculation
errors cannot be accumulated.
Entropy coding means 6 as shown in Fig. 1 compresses the single-bit
based bit map information stored by bit map information recording means 4
using a binary image coding such as run length coding, modified READ (MR,
MMR) coding, modified Huffman (MH) coding, or JBIG coding. More
specifically, in the case of a binary document image generally handled by a
facsimile machine or the like, there is a high probability that white pixels
or
black pixels successively appear in one or more continuous areas accordingly
with the run length coding method, a one-dimensional segment including only
white or black pixels, which are called run, is employed as a unit for coding,
and using the number of the continuous identical pixels included within each
run as the length of each run, encoding is performed. For example, in digital
facsimiles using the public telephone network, modified Huffman codes are
generally used for the run length model being established separately for black
and white pixels.
The modified Huffman coding (MH) is employed as a one-dimensional
coding method in facsimile transmission of monochrome pixel information
CA 02436437 2003-07-28
23
including 1728 pixels per scanning line, which is obtained by scanning at a
pixel density of, for example, 8 pixels/mm~ wherein the MH codes represent the
run lengths each being the length of each segment including only continuous
white pixels (white run) or only continuous black pixels (black run), and
variable length codes are assigned to the respective runs using the
statistical
tendency that white or black runs having particular lengths occur more
frequently than the other lengths, which is the theory for reducing the amount
of data.
The modified READ (MR or MMR) coding is used as a standard method
for two dimensional coding as well as one dimensional coding where the MMR
coding is the one that both the standard resolution and the high resolution
for
MR coding are set to infinity.
The basic structure of the compression method of motion image
information and system thereof according to the present invention is to
compare spatially or temporally adjacent pixels and output the resulting
difference information so that redundant information between frames can be
reduced. More specifically, pixel t and pixel t - 1 between frames are
sequentially compared, and information regarding on whether or not the
resulting difference is greater than parameter (threshold) P is stored as a
single bit piece of bit map information. Entropy coding means 6 predicts
codes that may occur within each frame and between frames, and outputs a
small error from the predicted value so that redundant information can be
reduced. It is well known that the average code length per pixel never be less
than or equal to the average information content (i.e., entropy) when codes
CA 02436437 2003-07-28
24
assignment is performed and the resulting sequence of codes are transmitted.
The adaptive Huffman coding algorithm is described below. The
adaptive Huffman coding is performed, so as to efficiently generate codewords
by collectively performing a series of predictive coding procedures including
S generating of difference information and Huffman coding thereof. According
to
the conventional Huffman coding, codewords are generally generated using a
Huffman table, and the generated Huffman table is updated whenever each
single word is encoded, or the Huffman coding procedure is dynamically
performed. In contrast, according to the adaptive Huffman coding, using as
many Huffman table (code table) as the predicted amount of information one of
that many tables is selected by a table selector in conformity with predicted
information with that selected table encoding is performed. Accordingly, data
such as audio information, which can be subjected to the predictive encoding,
is effectively compressed.
The adaptive arithmetic coding algorithm is described below. The
adaptive arithmetic coding is performed, so as to efficiently generate
codewords by collectively performing a series of predictive coding procedures
including generating of difference information and arithmetic coding thereof.
According to the conventional arithmetic coding, codewords are generally
generated using a single occurrence probability table, and the generated
occurrence probability table is updated whenever each single word is encoded,
or the arithmetic coding procedure is dynamically performed. In contrast,
according to the adaptive arithmetic coding, using as many arithmetic table
(code table) as the predicted amount of information one of that many tables is
CA 02436437 2003-07-28
selected by a table selector in conformity with predicted information with
that
selected table encoding is performed. Accordingly, data such as image
information, which can be subjected to the predictive encoding, is effectively
compressed.
S A specific structure of a predictive encoding circuit is shown in Fig. 3 in
which in order to encode input image data, which is analog to digital
converted,
it is suitably delayed and coupled to the table selector. The input image data
is
also transmitted without any delay to the encoding unit, which then encodes
it.
The resulting pieces of encoded data are compared and their difference is then
10 calculated. The table selector selects one code table for the input image
data in
conformity with predictive information, transmitting it to the encoding unit,
which in turn compresses the input image data so that adjusted codewords can
be obtained.
A specific structure of a predictive decoding circuit is shown in Fig. 4 in
15 which a codeword is transmitted to the decoder, and at the same time the
directly transmitted codeword is temporally sent to the table selector, which
then selects a decode table in conformity with predictive information sending
back it to the decoder, which in turn calculates difference from the
previously
decoded pixel value so that adjusted codewords can be obtained.
20 A method for compressing motion image information according to a
second embodiment of the present invention is further described.
The present invention is a method of compressing motion image
information by comparing spatially adjacent pixels within a frame or
comparing temporally adjacent pixels between frames, outputting the
CA 02436437 2003-07-28
26
resulting difference information of the pixel values, storing in a bit map,
information regarding on whether or not the output difference information is
greater than a given parameter (threshold) P, and compressing the difference
information that is greater than said parameter (threshold) P in conformity
with the information stored in said bit map thereby reducing redundant
information wherein the intra-frame compression is performed while changing
the divided block size within the same frame before the inter-frame
compression is performed.
According to intra-frame compression procedure, pixels within each
block are compared outputting the resulting difference information of the
pixels while changing the divided block size, and if that difference
information
is greater than parameter (threshold) P, a smaller block size is used for
portion
including this difference information.
In addition, if the difference information between pixels is greater than
parameter (threshold) P, a smaller block size is repeatedly used.
An image within a frame is divided into blocks, and each block is
approximated (substituted) with a single plane represented by at least three
components for pixels within each block.
Further, according to the intra-frame compression procedure, said plain
is represented by three pieces of data: the intensity of a pixel within a
block,
the gradient of the intensities within the block in the X direction, and the
gradient of the intensities within the block in the Y direction.
According to the intra-frame compression procedure, the entire image is
compressed in an n x m pixels block unit (n and m are integers, respectively)
CA 02436437 2003-07-28
27
using a intra-frame compression method, pixels between the original image
and the image expanded after compressed are compared outputting the
resulting each difference information for pixels, and if a pixel that caused a
larger difference than parameter (threshold) P to occur exists, the operation
of
using a smaller block size for a portion or a surrounding area including this
pixel is repeatedly performed until a designated minimum block size is
reached.
Further, if there is no change in the block size during the intra-frame
compression procedure, the inter-frame compression procedure starts.
If the block size is changed into a larger one, an additional calculation
for difference of data in said block is not performed outputting as it is.
If the block size is changed into a smaller one, difference from the
previous expanded data is calculated within each portion, and compressed in
that smaller block size.
Next, a system for compressing motion image information according to a
second embodiment of the present invention is described.
According to the intra-frame compression procedure, the entire image is
compressed in an n x m pixels block unit (n and m are integers, respectively)
using a intra-frame compression method, pixels between the original image
and the image expanded after compressed are compared outputting the
resulting each difference information for pixels, and if a pixel that caused a
larger difference than parameter (threshold) P to occur exists, the operation
of
using an ever smaller block size for a portion or a surrounding area including
this pixel is repeatedly performed until a designated minimum block size is
CA 02436437 2003-07-28
28
reached.
Further, with the block approximation means, if there is no change in
the block size during the intra-frame compression procedure, the inter-frame
compression procedure starts.
Further, with the block approximation means, if the block size is
changed into a larger one, an additional calculation for difference of data in
said block is not performed outputting as it is.
Further, with the block approximation means, if the block size is
changed into a smaller one, difference from the previous expanded data is
calculated within each portion, and compressed in that smaller block size.
In the following, a method and system for compressing motion image
information according to the second embodiment of the present invention are
described while referring Figs. 6, .7 and 10.
As described above, if a larger block size is used, compression ratio may
be enhanced however the detail of the original image may be lost deteriorating
image quality. A phenomenon where fine lines are completely lost happens
when the original image is configured by such fine lines the intensity thereof
differing from that of a fixed colored background. The following method is
utilized so as to solve these problems. To simplify explanation, the case (an
example) of 16 x 16 pixels of a white image is described.
More specifically, as shown in Figs. 6 and 10, the entire image is
compressed (expanded) in a 16 x 16 pixels block unit using an intra-frame
compression method as described above (STEP 1 in Fig. 10). Pixels between
the original image and the image expanded after compressed are compared
CA 02436437 2003-07-28
29
outputting the resulting difference information of each pixel, which is then
compared to parameter (threshold) P (STEP 2 in Fig. 10). As a result of this
comparison, if there is a pixel with difference exceeding parameter
(threshold)
P1, a portion or a 8 x 8 pixels block including this pixel is compressed
(expanded) (see STEP 3 in Fig. 10 and the largest circle in Fig. 6). In
addition,
the surrounding area of that portion is compressed in a 8 x 8 pixels block
unit.
Afterwards, pixels between the original image and the image expanded after
compressed are compared outputting the resulting difference information of
each pixel, which is then compared to parameter (threshold) P2 (STEP 4 in Fig.
10). As a result of this comparison, if there is a pixel with difference
exceeding
parameter (threshold) P2, a portion or a 4 x 4 pixels block including this
pixel
is compressed (expanded) (see STEP 5 in Fig. 10 and the middle circle in Fig.
6). In addition, the surrounding area of that portion is compressed in a 4 x 4
pixels block unit. Pixels between the original image and the image expanded
after compressed are compared outputting the resulting difference information
of each pixel, which is then compared to parameter (threshold) P3 (STEP 6 in
Fig. 10). As a result of this comparison, if there is a pixel with difference
exceeding parameter (threshold) P3, a portion or a 2 x 2 pixels block
including
this pixel is compressed (expanded) (see STEP 7 in Fig. 10 and the smallest
circle in Fig. 6). In addition, the surrounding area of that portion is
compressed
in a 2 x 2 pixels block unit. The procedure continues to the inter-frame
compression step (STEP 8 in Fig. 10). In this manner, it is possible to
compress
the original image with its detail maintained while maintaining a high
compression ratio.
CA 02436437 2003-07-28
Next, as a result of the intra-frame compression as mentioned above, an
inter-frame compression method in the case where a compressed image of
frame t in Fig. 7(A) and a compressed image of frame t + 1 in Fig. 7(B) are
obtained is described while referring Fig. 11. The block sizes are compared
5 (STEP 9)~ since there are no change in the block size between 1 in Fig. 7(A)
and 1' in Fig. 7(B) and between 2 in Fig. 7(A) and 2' in Fig. 7(B), difference
is
calculated and inter-frame compressed using one of the methods described in
the aforementioned inter-frame compression procedure (STEP 10).
Afterwards, it is determined whether or not difference in each block size is
10 equal to parameter (threshold) P or greater (in STEP 15). If it is
determined
that difference in each block size is equal to parameter (threshold) P or
greater,
the effect that THERE IS DIFERENCE is recorded in a bit map outputting the
difference (in STEP 16). Otherwise, if it is determined that difference in
each
block size is less than parameter (threshold) P, the bit map is stored and
15 updated with the effect that THERE IS NO DIFERENCE, outputting the
difference (see STEP 17). Incidentally, in the case where there is change in
the
block size, in particular where there is change towards more coarse resolution
(STEP 11) such as the change between 4 and 4' in Fig. 7, 4' is used as a key
block (or a key frame) that can be extended by itself independent from the
20 previous frame. In this case, no difference is calculated. In other words,
difference of data in the block 4' is not calculated outputting as it is (STEP
12).
In the case where there is change in the block size, in particular where there
is
change towards finer resolution (STEP 13) such as the change between 3 in
Fig. 7A and 3' in Fig. 7B, difference from the extended data within block 3 is
CA 02436437 2003-07-28
31
calculated for each portion and compressed in terms of the block size unit (in
STEP 14).
Next, a method of compressing motion image information according to
the third embodiment of the present invention is described.
According to the present invention, intra-frame encoded I frames (i.e.,
base frames solely from which the corresponding image can be reconstructed)
are used each I frame is spatially pre-divided into a plurality of I blocks,
which are then dispersed between each frame along the temporal axis.
When the respective I blocks, which are spatially divided, are dispersed
between each frame along the temporal axis, no I block is inserted into any
block within the frame updated at a time when a state that the difference
between frames is greater than parameter (threshold) P happens.
Furthermore, a compression method is provided where: an image within
a frame is pre-divided into blocks, all the divided blocks are each
approximated
(substituted) with a single plane represented by three pieces of data: the
intensity of a pixel in each block, the gradient of each block in the X
direction,
and the gradient of each block in the Y direction and using an intra-frame
coded I frame (i.e., a reference frame solely from which an image can be
reconstructed), said I fame is inserted within a sequence of frames wherein:
said I frame is spatially pre-divided into blocks when the divided I blocks
are
dispersed between each frame along the temporal axis, no I block is inserted
in
any block within the frame that has been updated due to difference between
frames being greater than parameter (threshold) P within a specific period of
time.
CA 02436437 2003-07-28
32
A method of compressing motion image information is provided, where
spatially adjacent pixels within a frame are compared or pixels between
temporarily adjacent frames are compared to output the resulting difference
information between the pixels information regarding on whether or not the
output difference information is greater than a given parameter (threshold) is
stored in a bit map and the difference information stored in the bit map that
is greater than said parameter (threshold) P is compressed thereby reducing
redundant information wherein: using an intra-frame coded I frame (i.e., a
reference frame solely from which an image can be reconstructed), said I fame
is spatially pre-divided into blocks and when the divided I blocks are
dispersed between each frame along the temporal axis, no I block is inserted
in
any block within the frame that has been updated due to difference between
frames being greater than parameter (threshold) P within a specific period of
time.
Further, a method for compressing motion image information according
to the third embodiment of the present invention is described.
According to the present invention, intra-frame encoded I frames (i.e.,
base frames solely from which the corresponding image can be reconstructed)
are used each I frame is spatially pre-divided into a plurality of I blocks,
which are then dispersed between each frame along the temporal axis.
I block insertion means 7 does not insert any I block in a block within
the frame that has been updated due to difference between frames being
greater than parameter (threshold) P within a specific period of time.
Furthermore, a compression system is provided comprising a block
CA 02436437 2003-07-28
33
approximation means for dividing an image within a frame into blocks and
approximating (substituting) each of all the divided blocks by a single plane
represented by three pieces of data: the intensity of a pixel in each block,
the
gradient of each block in the X direction, and the gradient of each block in
the
Y direction said system further comprises: I block generation means 8 for
spatially dividing an intra-frame coded I frame (i.e., a reference frame
solely
from which an image can be reconstructed) into I blocks and I block insertion
means 7 for inserting an I block in a portion except for a block within the
frame that has been updated due to difference between frames being greater
than parameter (threshold) P within a specific period of time, when the
divided
I blocks are dispersed between each frame along the temporal axis.
In the following, a method and system for compressing motion image
information according to the third embodiment of the present invention are
described while referring Figs. 8, .9 and 12.
The present invention is an encoding method, which corresponds to
partial sequence image reconstruction (decoding) and/or data error occurring
during that image reconstruction. It is noted that there is a premise that a
compression algorithm is used without use of any motion prediction and
correction technique for more than three entire frames that are to be
compressed.
First, as shown in Fig. 12, an intra-frame predictive coded frame or an I
frame is spatially divided into a single or multiple I blocks (in STEP 1), and
these divided I blocks are dispersed along the temporal axis (generation of I
blocks in STEP 2.) It is noted that the block size, the divided block shape,
etc.
CA 02436437 2003-07-28
34
due to that generation of I blocks can be optionally changed, and moreover
they can be randomly selected.
More specifically, as shown in Fig. 8, an I frame of 8 x 8 pixels is
spatially divided into sixteen I blocks each having 2 x 2 pixels, and these
are
inserted into a sequence of frames at fixed period intervals. As a result,
when
the portion where a inter-frame difference output occurs (i.e., the portion
with
large information content where there is a motion) and an I block (the
information content of which is greater than that of the other frame)
overlaps,
a useless I block should be inserted drastically increasing the information
content, which may cause errors impossible to be recovered to happen within
the inserted I frame. In order to avoid this, as shown in Fig. 12, in cases
where
there is no problem with the processing speed on the encoding side, whether or
not the state of the inter-frame difference greater than parameter (threshold)
P occurs within a designated period of time and updating (outputting of the
difference) is accordingly performed is determined (in STEP 3), and no I block
is inserted in any block that is updated (or that outputs a difference) (in
STEP
4.) On the other hand, an I block is inserted to the block that is not updated
(or
that outputs a difference) (in STEP 5.)
Referencing Fig. 9, a specific encoding method is described. It is noted
here that an I frame of 8 x 8 pixels is spatially divided into a block of 1 x
2
pixels by I block generation means 8 configuring thirty-two I blocks
altogether,
as an example. It is also noted here that a motion image is provided as an
example comprising an image frame with a block of 8 x 8 pixels and with the
maximum block of 16 x 16 pixels. In Fig. 9, as a matter of convenience, the (n
+
CA 02436437 2003-07-28
11)th to the (n + 32)th are omitted.
First, the I blocks each horizontally having 1 x 2 pixels, which are
marked in black in the Figure, are inserted. An object (dark gray regions
representing a difference outputting block that moves against the background),
5 which is initially positioned on the upper left corner in the image and
which
requires the maximum of 2 x 2 pixels to be updated (i.e., to output
difference)
moves towards the lower left. Until the (n + 3)th frame, an I block is
generally
inserted (in STEP 5 of Fig. 12.) In contrast, since the block corresponding to
the object appeared on the upper right in the (n + 3)th frame is updated
(i.e.,
10 outputs difference), the I block to be inserted within the (n + 4)th frame
is not
actually inserted (see the shaded portion and STEP 4 in Fig. 12). It is noted
that the light gray portion denotes a difference outputting block that is
coming
back to portion of the original background because the object has moved. In
this case, the processing where no I block is inserted (see STEP 4 in Fig. 12)
15 appears in the (n + 7)th frame and (n + 8)th frame. More specifically, as a
result of the movement of the object in the (n + 5)th frame, if a difference
outputting block (the light gray portion) that should be back to portion of
the
background exists as a portion to be updated, within the (n + 7)th frame, only
the single block on the right side of the I blocks each having 1 x 2 pixels is
not
20 inserted in that portion. As a result of the movement of the object in the
(n +
4)th and the (n + 5)th frames, if a difference outputting block (the light
gray
portion) of horizontally positioned 1 x 2 pixels that should be back to
portion of
the background exists as a portion to be updated, within the (n + 8)th frame,
an I block of 1 x 2 pixels is not inserted in that portion. In this case, the
CA 02436437 2003-07-28
36
reference time (the near past) during which any I block should not be inserted
is represented by the number of the frames needed for I blocks to be inserted
to
every block position (8 x 8 / 2 = 32 frames.) In other words, no I block is
inserted in any blocks that are updated (outputs difference) within a sequence
of thirty-two frames due to the movement of an object, etc. In order to start
reconstructing a desired frame at an arbitrary temporal position, decoding
should begin a predetermined number of frames earlier to allow complete
reconstruction of a single image.
INDUSTRIAL APPLICABILITY
As described above, with a motion image information compression
method and system thereof, according to the first embodiment of the present
invention, an intra-frame image is pre-divided into blocks, every divided
block
is approximated with a single plane represented by three pieces of data: the
intensity of a pixel within each block, the gradient of said each block in the
X
direction, and the gradient of said each block in the Y direction, thereby
efficiently performing the intra-frame compression.
According to a motion image information compression method and
system thereof, the intra-frame compression is performed by compressing the
entire image in an n x m pixels block unit (n and m are integers,
respectively),
pixels between the original image and the image expanded after compressed
are compared outputting the resulting difference information of each pixel,
and
if a pixel that caused a larger difference than parameter (threshold) P to
occur
exists, repeatedly using a larger block size for a portion or a surrounding
area
CA 02436437 2003-07-28
37
including this pixel until a designated minimum block size is reached, thereby
maintaining the detail of the original image and preventing deterioration of
image quality.
With a motion image information compression method and system
thereof, according to the third embodiment of the present invention, an I
frame
is spatially pre-divided into I blocks, and when the divided I blocks are
dispersed between each frame along the temporal axis, no I block is inserted
in
any block within the frame that has been updated due to difference between
frames being greater than a given parameter (threshold) within a specific
period of time therefore reconstruction of an image can be performed by
starting reconstruction a predetermined number of frames earlier so that a
single image can be completely reconstructed, and displaying the reconstructed
image after the temporally positioned target frame is reached thereby easily
displaying a reconstructed image in an optionally temporal position without
much time being taken for searching for the I frame.
In addition, since the amount of distributed data at a delivery server
and/or in the data communication path during the delivery of a motion image
is temporally uniformized, a higher distribution performance than that with
the conventional content distribution technique is obtained. On the reception
/
reconstruction side, since change in the received amount per unit time is
small,
a necessary amount of buffering memory can be reduced, and moreover since
the burden imposed on the reconstruction processing is regulated, even a
system with low performance can perform steady reconstruction. In addition,
since the influence of data errors on the reconstruction processing is small,
the
CA 02436437 2003-07-28
38
reconstruction processing can continue with such data errors neglected
therefore, it is unnecessary for the distribution side system to re-send data,
and lighter burden is imposed on the distribution side. Moreover, it is also
possible to easily provide multicast distribution capability, etc. for the
motion
image broadcasting.
As described above, the present invention is an optimal means for
efficiently compressing motion image information, and can be widely used in
the fields of transmission, reception, and reconstruction of a variety of
motion
image information.
