Note: Descriptions are shown in the official language in which they were submitted.
35-DV-123
~31'76~
The present invention relates to a solid state imaging
system and more particularly to solid state imaging systems
designed to obtain video color signals. The invention
further relates to color filtering and electrical signal
processing necessary to obtain a three color electrical
video signal.
It has been proposed to use solid state imagers for
color. A rather direct approach would be to use two or
three imagers deriving an image from a common scene through
separate filters. The cost of solid state imagers is
sufficiently high, however, to reserve this approach for
applications where the expense of an added imager is per-
missible. In low cost applications, it is desirable to
use a single imager. Such a system has been proposed in
U.S. Patent No. 3,890,500 dated June 17, 1975 entitled
"Apparatus for Sensing Radiation and Providing Electrical
Readout" to Eichelberger et al, and assigned to the present
Assignee. The color system there described employs a
single imager using striped filters. Individual strips are
aligned with individual sensor columns in groups of two or
three. In a three color system, for instance, a red pass,
a green pass and a blue pass color filter stripe are
arranged in repeating sequence of three. The imager is
then addressed to the red lines in one field, the green lines
in a second field, and the blue lines in a third field in
repeated sequences to obtain a three color signal. The
above approach is of somewhat lower resolution than optimum
and is field sequential.
Accordingly, it is an object of the present invention to
provide an improved solid state color imaging system.
It is a further object of the present invention to
provide an improved solid state color imaging system employing
35-DV-123
~3~7~
a single solid state imager.
It is an additional object of the present invention to
provide a solid state color imaging system employing a single
solid state imager and achieving improved resolution.
These and other objectives of the invention are achieved
in a novel solid state color camera comprising a color
filter, a solid state imager, and a color processor network.
The color filter has a first succession of stripes aligned
parallel to a vertical dimension and spaced at a fixed
interval, the stripes alternately passing and rejecting a
first primary color. Typically, the passing stripe is clear
and the first color is red. The solid state imager contains
an area array of image sensors, arranged in rows and columns.
The columns of sensors are aligned parallel to the vertical
filter dimension at the fixed filter stripe interval, and in
registry with the filter stripes. Output means are provided
coupled to the array for deriving signals from the individual
sensors at a given periodic rate. The color processor
network obtains an electrical signal corresponding to the
above-mentioned primary color. It comprises a network for
subtractively combining the electrical signal from a sensor
beneath the first color passing stripe with the electrical
signal from an adjacent sensor beneath the first color
rejection stripe. The adjacent sensor substrations are
repeated throughout each row to produce a first primary
color signal.
In accordance with one facet of the invention, the
signals are derived sequential by the output means from
the sensor elements two elements at a time, signals from
odd sensor elements being coupled to a first output terminal
and signals from even sensor elements being coupled to a
second output terminal. The color processor comprises a
35-DV-123
76~
differential amplifier having one input coupled to the first
output terminal and a second input coupled to the second
output terminal, the output corresponding to the first color
appearing at the output of said differential amplifier.
Instrumental to selecting a second (and third) color
signal, the color filter has a second succession of stripes
aligned parallel to the vertical dimension and spaced at
double the fixed interval of the first succession. The
second succession of stripes alternately passes and rejects
a second primary color and forms, together with the first
succession of stripes, a composite color filter having a
four stripe color differentiated, repeating sequence. In
the second succession, the second primary color passing
stripe is clear and the second primary color rejecting stripe
rejects blue. In the color processor, the second primary
color signal is obtained by means for additively combining
the electrical signal from the first odd and even sensors
beneath the first two stripes of each four stripe sequence
to form a first value and additively combining the electrical
signal from the second odd and even sensors beneath the
second two stripes of each four stripe sequence to form
a second value. The two additions are repeated in each
four stripe sequence to form an a.c. wave whose amplitude
is equal to the difference between these values.
In accordance with another facet of the invention,
assuming that signals are derived from odd sensor elements
at a first output terminal and signals from even sensor
elements at a second output terminal, the additive means of
the color processor are coupled respectively to the first
and second output terminals of the output means. Delay
means are then provided coupled to the output of the
additive means to delay the signal from each pair of sensors
35-DV-123
1~31'76~
into time coincidence with the later signal from the adjacent
following pair of sensors. Means are then provided for
subtractively combining the delayed and undelayed signal to form
a quantity corresponding to said second color. The quantity
so formed ~n an ac quantity. Means are then provided for
detecting the amplitude of the ac wave to obtain a voltage
proportional to the second primary color.
Assuming that the first and second colors are red and
blue, one position in each four stripe sequence, rejects red
and blue, and effectively passes green. The color processor
has a network for obtaining an electrical signal for the
green primary color by sampling means, timed to selected
only the electrical signal from a sensor beneath the green
pass position and repeating this selection in each four
stripe sequence.
The processor also includes a network for obtaining
a luminance signal. This network comprises means for
additively combining the electrical signal from a sensor
beneath a clear stripe with the electrical signal from a
sensor beneath a blue rejecting stripe in the same four
stripe sequence, one of the two signals being delayed prior
to combination by one horizontal element scanning period.
The network also comprises means for additively combining
the electrical signal from a sensor beneath a blue re-
jecting stripe with the signal from a sensor beneath a
blue and red rejecting stripe in the same four stripe
sequence, one of the two being delayed prior to combination
by one horizontal element scanning period. For both delays,
the delay means is provided in either the path from the odd
or the even sensor output terminal. The two additive
combinations are repeated in each four stripe sequence to
form a first luminance signal.
~ 7~ 35-DV-123
In accordance with other aspects of the invention, the
processor is provided with a network for correction of non-
blue transients in the network for obtaining a blue channel
and a network is pxovided for correction of an undesired
blue stripe modulation in the network for obtaining a
luminance signal. The color processor is completed by
three low pass filters provided respectively for the first,
second and third color signals having a common higher fre-
quency cut off and a filter for the luminance signal whose
lower cut off frequency maches the upper cut off frequency
of the first three filters. Three adders are provided for
adding the luminance signal to each of the color signals.
The final three color signal is available at the outputs
of these three adders.
The novel and distinctive features of the invention
are set forth in the claims appended to the present
application. The invention itself, however, together with
further objects and advantages thereof may best be under-
stood by reference to the following description and
aceompanying drawings, in whieh:
Figure lA is a side elevation view of the filter
and the solid state imager whieh are used in the novel
eolor imaging system herein deseribed;
Figure lB is a plan view of a filter and solid state
imager together with a bloek diagram of a color processor
for separating the signals from the imager into electrical
signals representing the primary color components;
Figure 2A is a drawing illustrating the operation
of the red channel of the color imaging system in con-
sc~
verting light from a red s~ into an electrical signal;
Figure 2B is a drawing illustrating the operation ofthe blue channel of the color imaging system in converting
-- 5 --
1~176~ 35-DV-123
a blue scene into an electrical signal;
Figure 2C shows the operation of the blue color circuitry
in treating the blue scene;
Figure 2D is an illustration of the creation of undesired
transients in the blue channel in treating a none-blue signal
and their correction in a separate network;
Figure 2E is a drawing illustrating the creation of
undesired blue stripe modulation in the luminance channel and
its correction in a separate network; and
Figures 3A and 3s form an electrical circuit drawing of
a practical embodiment of the invention omitting those portions of
the color processor circuitry prior to the final filtering and
the addition to each primary color of wide band luminance
information.
Referring now to Figures lA and lB, a color imaging syste~
using a single solid state imager is shown. Figure lA is a side
elevation view of the filter and the imager. As seen in Figure lA,
light from an object is passed through a composite color filter
11 prior to impingement on the surface of the solid state CID
(charge injection device) imager 12 upon which an image is focused.
Figure lB is a plan view of the imager, together with a block
diagram of a color processor. As seen in Figure lB, the imager
12 contains a rectangular array of image sensors 10. The imager is
coupled to a scanning generator (not shown) which controls the
conversion of the information on the sensor array into a pair of
time variant electrical signals (eO, ee) at two imager output
terminals 13 and 14. The processor to which the signals from the
imager are applied is shown in block diagram form and comprises
the blocks 31-35. The function of the processor is to separate
the signals from the imager into three color components, typically
R, G and B.
The composite color filter 11 is illustrated in both
Figures lA and lB. As seen in Figure lA, the filter is
- 6 -
~ . 76~ 35-DV-123
seen to consist of a supportive glass substrate 15 upon the
undersurface of which are formed two layers 16 and 17.
These layers are shown somewhat artifically as being of
uniform thickness and with a thickness in relation to the
thickness of the glass substrate which is much greater than
in actuality. The first layer 16 consists of an alternating
series of clear (marked w) and red excluding (marked -R)
strips. The other (or lower) layer 17 consists of an alternat-
ing series of clear (marked w) and blue excluding (marked-B)
strips at one half the spatial frequency of the upper layer.
The composite effect of the two layers is to form a filter
(11) in which the filter function occurs in recurring sets
of four; one member of the set being clear (w), one member
of the set excluding red(-R), one member of the set exclud-
ing blue (-B), and one member of the set excluding red and
blue -R, -B) and approximating a filter passing green (+G).
The alignment of the composite color filter in respect to
the array of optical sensors on the imager, is best under-
stood in connection with the plan view of the imager.
The CID imager 12 includes vertical (18) and horizontal
(19) shift registers and vertical (20) and horizontal (21)
enable gates which assist in scanning the image focused on
the sensor array. In a typical arrangement,the imager
consists of an array of optical sensors with 244 in the
vertical dimension by 320 in the horizontal dimension,
all integrated on a common substxate. Each sensor element
10 in the array has two terminals, one of which is associated
with vertical selection and the other of which is associated
with horizontal selection. The "vertical" terminals are
interconnected in buses which extend in the horizontal
dimension, all integrated on a common substrate. Each
sensor element 10 in the array has two terminals, one of
35-DV-123
7~
which is associated with vertical selection and the other
of which is associated with horizontal selection. The
"vertical" terminals are interconnected in buses which
extend in the horizontal dimension and which are spaced
along the vertical dimension. These buses are called "rows".
The horizontal terminals are interconnected in buses which
extend in the vertical dimension and which are spaced along
the horizontal dimension. These buses are called "columns".
The rows are selected one row at a time by a vertical
enabling pulse acting in conjunction with a vertical pulse
upon a two input, enable gate (20) (also integrated). The
vertical shift register 18, which is integrated on the
common substrate, has 122 stages, each of which is coupled
to a ~d of enable gates (20). An enable pulse, which
lasts for a "field" turns "on" the first input of the
first gate of all pairs of vertical enable gates. As the
vertical pulse is propagated along the vertical shift
register, its presence at each stage in conjunction with
the enabling pulse turns on the first of each pair of
vertical enable gates. When all "odd" rows have been
sampled to complete one field, a second vertical enable
pulse turns on the second of each pair of vertical enable
Co ~ ~ ~\5e
gates. As the next v~-ical~se is propagated down the
vertical shift register 18, the second of each pair of
vertical enable gates (20) is turned on. The result is a
selection of odd sensor rows in a first field and even
rows in the second field.
Once a row has been selected at the slower vertical
rate, it is scanned rapidly at the faster horizontal rate.
The sensor elements are arranged in the horizontal dimension
in 160 odd and 160 even "columns". The columns are
selected one pair at a time by a horizontal pulse propagated
-- 8
35-DV-123
~3~76~1
along 160 stage shift register 19, also integrated on the
common substrate. Each stage of the horizontal shift re-
gister is coupled to a pair of horizontal enable gates (21)
(also integrated). Each horizontal enable gate controls
a column of sensors, the odd gate of the pair coupling the
sensor element to an odd (eO) video signal output and the
even gate of the pair coupling the sensor element to an even
(ee) video signal output. As the horizontal pulse is pro-
pagated along the horizontal register, consecutive adjacent
pairs of sensor elements are "enabled", producing a simul-
taneous pair of odd (eO) and even (ee) video output signals
at the respective output terminals of the array.
The sensor array is scanned in the following manner.
The scanning generator, acting through the vertical shift
register (18), selects a pair of rows. The first row in
that pair and the following pairs is activated by a vertical
enabling pulse, and the horizontal shift register (19)
steps along the row, two adjacent horizontal elements at a
time. In succession, the first row of the second pair of
rows, i.e. row 3, which is also activated by the enabling
pulse, is scanned followed in succession in each of the
odd rows until the first field is completed. After the
first field is completed, the vertical enabling pulse now
enables the second row of the first pair of rows and the
second row of all the following pairs of rows. In succession,
the second, fourth and all the even rows are scanned until
the second field is completed. The process of scanning the
odd rows is then repeated for the third field, etc.
As previously noted, the filter 11 is critically aligned
in respect to the image sensors 10 of the imager 12 as shown
in the plan view of Figure lB. The light sensitive areas
of the optical sensors are shown as small squares typically
~ 35-DV-123
~3 !l76~
spaced on centers at .0012 inch intervals. The light
sensitive area of the array, assuming a 320 column by
244 row array, is approximately 0.4 inches by 0.3 inches.
The long dimension of the filter strips are arranged parallel
to and in precise registry with the sensor columns. In
particular, the first clear (or w) strip of the filter is
superimposed upon the leftmost column of sensors. The
second red rejecting strip of the filter is superimposed
upon the second column from the left. The third, blue re-
jecting strip of the filter is superimposed upon the thirdcolumn of sensors. The fourth, or green, passing strip of
the filter is superimposed upon the fourth column of sensors
from the left. The filter contains a repeating succession
of strips in similarly ordered sets of four throughout
the remainder of the 320 columns.
It is normally preferably for the filter to be placed
in close proximity to the sensor array to avoid the need
for any additional lenses. The camara lens, which is
not shown, may have a numerical aperture between 1 and 2,
leading to a focal length of less than one inch with a
sensor array of the indicated size. With a focal length
this short, the filter must be in close proximity to the
sensor elements for color accuracy. Close proximity is
achieved by placing the filter layers 16 and 17 on the
undersurface of the filter substrate, and attaching the
undersurface of the filter 11 directly to the upper surface
of the CID imager. Alternatively, the filter may be formed
on the upper surface of the CID imager.
The color processor for converting the imager output
into separate electrical sigr,als for the R, B and G color
components is shown in both the block aiagram of Figure lB
and the detailed circuit diagram of Figure 3. The color
.;
-- 10 --
.
1~31~6~ 35-DV-123
processor, as will be explained, cooperates with the striped
filter in deriving the individual electrical color signals.
The red information is contained in alternate sensors in
which the "clear" sensors are alternated with those beneath
a red rejecting stripe. This filtering is set by filter
layer 16. The stripe system limits the bandwidth available
to red information to a spatial frequency of one-half the
sensor interval. The blue information is contained in sets
of four sensors in which two adjacent sensors beneath the
clear band of filter layer 17 act in alternation with two
s~c ~
adjacent ccnspls beneath the blue rejecting stripe of filter
layer 17. The stripe interval of filter layer 17 limits
the blue information bandwidth to a spatial frequency of
one-quarter the sensor interval. Green and approximate
luminance information on the other hand is available at
the outputs of all sensors, and the bandwidth is not limited
by the periodicity of the stripes. In obtaining the green
signal a sample is used, however, which makes a sample once
in every four stripes, and thus limits the bandwidth of the
green information to the same bandwidth as the blue. In
practice, the bandwidth available to each color component
as a function of this stripe and the decoding process is
restricted to 0.5 megahertz by a following electrical
filter. The reduced bandwidth in the individual color
channels produces ~ligablc net loss in system re-
solution since high frequency luminance (YHF) infomration
is re-introduced into each color channel in frequencies
above those used in the individual color channels. The re-
introduction is made in appropriate proportion to maintain
its true nature as luminance information (Y) as a final
stage in formation of the individual color signals (Y =
0.59G + 0.3R + O.llB).
-- 11 --
1~ 3~764 35-DV-123
Referring again to Figure lB for a consideration of the
blocks of the color processors, the outputs (eO,ee) from the
imager are applied to an odd (31) and an even ~32) video
preamplifier which are the input elements of the processor.
The preamplifiers 31 and 32 are normally accompanied by
noise cancellers, low pass filters, and sampling circuitry
(not illustrated) designed to reduce pattern noise and to
increase the signal to noise ratio. The output from the
preamplifiers 31, 32 and any other input circuitry is then
distributed to four (generally separate) channels including
three color channels and a luminance channel. In the
illustrated embodiment, the blue and luminance channels are
twice inter-connected to provide two corrections, and finally
the luminance is re-introduced into each color channel as a
last stage in the processing to form the output into a
conventional three color format.
The electrical signal corresponding to the red color
component is derived from the outputs of the odd and even
video preamplifiers (31 and 32) by the blocks 33, 34 forming
the red color channel. The operation of these blocks in
treating a red color bar is illustrated in Figure 2A. Re-
ferring initially to Figure lB, the block 33 is a differen-
tial amplifier having one positive (+) input coupled to the
output of the odd video preamplifier 31, and the other, or
negative (-) input coupled to the output of the even video
preamplifier 32. The difference (eO ~ ee) is then passed
through a low pass filter 34, having an upper limit of
0.5 megahertz, and then applied to one input terminal of the
adder 35. The luminance signal, whose derivation will be
described subsequently, is applied to the other terminal of
the adder 35. A red color signal having a bandwidth ap-
preciably enhanced by the addition of the luminance in-
~ 6~ 35-DV-123
formation is available at the output terminal of the adder
35.
In short, the electrical signal corresponding to the
red color component is obtained in the red channel by a sub-
traction of the even (ee) from the odd (eO) imager signals
to obtain a low resolution red signal followed by the
addition of a luminance signal to obtain a high resolution
red signal. An understanding of the first stage of this
operation may be gathered from a consideration of Figure 2A.
In Figure 2A, a red scene is illustrated above the strip
filter layers 16 and 17. The red scene is focused through
the filter upon the sensors in columns numbered 5 through
12. The output (H~) of the horizontal shift register is
shown, producing pulses 3 through 6 in coincidence with the
red scene. Under these circumstances, the odd imager signal,
corresponding to that derivative from sensor elements 5,
7, 9 and 11, products a (1) output for each red illuminated
element. Taking up each of these four odd sensor elements
separately: in sensor element 5, the red light from the
object passes through the clear sections of filter layers
16 and 17 to produce a first "1" output; in sensor element
7, the red light passes through the clear section of filter
layer 16 and the blue trap of filter layer 17 to produce a
"1" output; in sensor element 9, the red light passes
through two clear sections of the filter layers to
produce a "1" output; and in sensor element 11, the red
light passes through a clear and blue trap section of
the filter layers to produce a "1" output. The odd sensor
output for the same red scene is shown at eO. Under these
circumstances, the even column imager signal and in parti-
cular sensor elements 6, 8, 10 and 12 within the red scene
13
35-DV-123
1~3~7~
produce no output. In sensor element 6, the red light
from the object impinges on the red trap of the
first filter layer 16, and produces a zero output. The
red traps in the filter layer 16 at the eighth,
tenth and twelfth sensor positions produce a similar zero
output. The even sensor waveform for a red scene is
shown at ee in Figure 2A. When the two waveforms are
subtracted in differential amplifier 33, the red signal
(eR) illustrated in Figure ~A is produced.
The foregoing process, by which the outputs of two
adjacent elements are subtractively combined to obtain
one sample of red information, in effect sets a requirement
of two adjacent sensor elements for each red picture element.
The maximum available spatial frequency and thus the band-
width for the red information is one-half the sensor interval.
The electrical signal corresponding to the green
component is derived from the output of the even video pre-
amplifier 32 by the blocks 36, 37, 38 and 39 forming the
green signal. The operation of these blocks in treating a
green color bar is illustrated in Figure 2B. Referring
initially to Figure lB, the block 36 is a sample and hold
circuit whose input is coupled to the output of preamplifier
32. The sampling function is controlled by a green sampling
waveform (esg) at a frequency which is half the horizontal
clocking frequency. The hold function is designed to
provide a sustained output between sampling pulses. As will
be explained below, the sampling pulse is timed to select
the output from the fourth, or green pass section of each
four strip filter set. The output of the old circuit is
supplied to a buffer amplifier 37, passed through a low
pass filter of 0.5 megahertz to one terminal of the adder
B -14 -
35-DV-123
76~
39. The luminance signal is applied to the other terminal
of the adder 39. A high resolution green color signal, as
a result of the addition of the luminance information, is
available at the output terminal of the adder 39.
As noted above, the electrical signal corresponding
to the green component is obtained in the green channel by
sampling the even (ee) imager signal to obtain a low re-
solution green signal followed by the addition of a luminance
signal to obtain a broadband green signal. A green scene
as illustrated in Figure 2s may be referred to to explain
the stage of this operation. A green scene is illustrated
in Figure 2B above the strip filter layers 16 and 17. The
green scene is focused through the filter upon the image
sensors numbered 5 through 12. The output of the horizontal
shift register (H~) is shown producing pulses 3 through 6 in
coincidence with the green scene. As described earlier,
the super-position of a red rejection (-R) and a blue rejec-
tion (-B) filter produces a green pass (+G) filter. The
green pass occurs in the fourth element of each four strip
set and thus occurs once every four elements. In the
example, the filter produces a green pass indicating a "1"
output at sensor positions 8 and 12. Signals indicating
illumination would also be produced, from illumination,
whether green or not, falling on the clear sections of each
filter set, corresponding to sensor positions 5 and 9.
Similarly, "1" output signals would be produced at the
sensor positions 6 and 10, corresponding to the red trap,
whether blue or green light were applied. Finally, light
indicating output signals would be produced at the sensor
positions 7 and 11 corresponding to the blue trap, whether
red or green light was applied. To select only the green
information, only the fourth section of each filter set is
- 15 -
1~3~64 35-DV-123
read. The green is selected by a sampling and hold circuit
36 controlled by a sampling pulse (es ) As illustrated,
the green sampling pulse (es ) is designed to sample at
half the horizontal clocking rate, corresponding to every
fourth sensor element, and in particular to sample the
alternate even sensor elements 4, 8, 12 and 16 under the
green pass filters. Assuming a green scene at sensor elements
5 through 12, samples will occur at sensor elements 8 and 12.
With the indicated four element holding period, the output
waveform eg appears at the output of the hold portion of
the circuit at a time corresponding to sensor elements 8 - 15.
The green output in this example is delayed some three elements.
The green information, while available at all four stripes
in each four stripe set, is exclusively available in only
the fourth stripe of each set. The process of taking a
sample every fourth stripe to obtain the narrow band green
information thus derives information at a spatial frequency
equal to one-fourth the spatial frequency of the sensors.
The low resolution green (eg~, after buffering in element
37 and filtering in 38 is then combined with the luminance
signal.
The electrical signal corresponding to the blue color
component is derived from the outputs of the odd and even
video preamplifiers (31 and 32) by the blocks 40 to 47 and
55 forming the blue channel. The operation of these blocks
in treating a blue scene is illustrated in Figure 2C. The
creation of a false blue transient from non-blue light, and
its correction by blocks 45 and 46 is illustrated in Figure
2D. Treatment of non-blue light will be deferred. As shown
in Figure lB, the outputs of the video preamplifiers are
coupled to the separate inputs of the adder 40. The adder
output is then seprated into a first part, which is applied
- 16 -
" 35-DV-123
L76~
to the two sensor element delay (320 nanosecond) 41, and an
undelayed second part. The delayed and undelayed outputs
are then applied respectively to the positive and negative
input terminals of the differential amplifier 42. The
difference from 42 is then coupled to one input of an adder
55 to the other input of which a blue correction signal
from the luminance channel via blocks 45 and 46 is added.
In the blue channel, one picture element of blue information
is available in each set of four sensors, assuming the signal
derivation process noted. The sum appearing at the output
of 55 is coupled to the absolute value amplifier 43. The
absolute value amplifier,which is a wide band rectifier,
recovers the peak amplitude of an applied ac waveform. The
rectified output from 43 is filtered in low pass filter 44
and supplied to one terminal of the blue adder 47. The
luminance signal is added to the other terminal of the
adder 47. A high resolution blue color signal as a result
of the addition of the luminance information of the low
resolution blue signal is available at the output terminal
of the adder 47.
The operation of the blue color circuitry in treating
the blue scene is illustrated in Figure 2C. As seen in
Figure 2C, the blue scene is focused through the filter 11
upon the sensors 5 through 14 corresponding to timing
intervals (H~ ) 3 to 7. The odd imager signal (eO), cor-
responding to that derived from sensor elements 5, 7, 9,
11 and 13 produces a l+O-~l+O+l output. In particular, in
sensor element 5, the blue light passes through the clear
sections of filter layers 16 and 17 to produce a "1" output.
In sensor element 7, the blue light passes through the clear
section of filter layer 16, but is trapped in the blue
trap of filter layer 17 to produce a "O" output. In sensor
- 17 -
~1~1764 35-DV-123
element 9, the blue light passes through two clear sections
of 16 and 17 to produce a "1" output, and in 11, the blue
trap in layer 17 produces a "0" output. In sensor element
13, the sequence begins again with +1. The illustrated eO -
waveform is the net result. Under the same circumstances,
the even imager signal (ee), corresponding to that derived
from sensor elements 6, 8, 10, 12 and also produces a
1+0+1+0+1 output. Adding eO to ee in the adder 40 produces
the 2+0+2+0+2 waveform labelled (eO + ee) The neYt wave-
form (eO + ee) d illustrates the effect of a two element
delay in block 41. The delayed (eO + ee) and delayed
(eO+ ee)d waveforms are then subtracted in 42 to ontain a
(+2-2+2-2+2-2) ac waveform. After addition of a blue
correction in adder 55, this ac waveform is then applied
to the absolute value amplifier 43. The absolute value
amplifier 43 produces a dc value of twice the original signal,
corresponding to the illustrated double amplitude "ac"
excursions. The dc output is filtered in 44 and combined
with the luminance in blue adder 47 to produce a higher re-
solution blue signal.
The luminance channel and its operation will now be
discussed, deferring until last, a consideration of the
two correction networks, which supply corrections to both
the blue and the luminance channels.
The electrical signal corresponding to the luminance
component is derived from the outputs of the odd and even
video preamplifiers (31 and 32) by the blocks 48 to 54
forming the luminance channel. The correction circuitry
for the luminance channel comprises the blocks 48 and 50.
The operation of the luminance channel is not specifically
illustrated, except in respect to the correction of blue
stripe interference, which i5 illustrated in Figure 2E.
- 18 -
35-DV-123
6~
As shown in Figure lB, the output of the odd video pre-
amplifier 31 is coupled through the adder 48, to an input
terminal of the adder 51. The output of the even video
preamplifier 32 is coupled successively through an adder
50 and through the one sensor element delay (160 nanoseconds)
49 to a second input terminal of the adder 51. The output
of the adder 51 is coupled to a first input terminal of
the luminance adder 53. The even sensor output, which is
delayed in delay element 49, is also coupled to a first pair
of input terminals of a switching differential amplifier 52.
The odd video preamplifier output at the output of the adder
50 is coupled to the other pair of input terminals of the
switching differential amplifier 52. The switching wave-
form applied to the differential amplifier is designed to
provide output inversion at the sensor element rate (i.e.,
sensor element 1 is +; 2-;3+;4 -; etc.). The outputs of
blocks 51 and 52 are then combined at the input terminals
of the luminance adder 53. (Also coupled to the output of
the block 51 is an input connection to the blue transient
correction circuitry 45, 46.) A high resolution luminance
signal, subject to blue stripe interference, appears at the
output of the adder 53. The luminance signal is then
coupled to a filter 54 which passes only luminance information
between 0.5 and 3.5 megahertz. The filtered output is
coupled to the inputs of each of the adders 39, 35 and 47
to form the high resolution G.R and B color signals.
The luminance channel, (neglecting error correction),
uses a known branched circuit for preserving the bandwidth
of the luminance information in combining the odd and
even video signals. The adder 51, in combining the delayed
even and undelayed odd signal, blurs transitions and in
effect attenuates the higher frequency information in the
-- 19 --
~3~76~ 35-DV-123
video signal. It produces the low frequency luminance
information (YLF). This "blurring" may be explained as
follows: Let us assume that a transition or picture edge,
for instance going from black to white, occurs at some
element in the array. The odd channel signal eO produces a
step between a pair of array sensors at this transitivn.
Similiary, the even channel signal e produces a step.
Because of the delay in the delay element 49, the even channel
signal shows the same step delayed one horizontal element.
The result is a sloping, two step transition as opposed to
a steep, single step transition. It is evident that the
two steps can not be coincident because of the delay of one
signal in relation to the other. The same effect is present
when the reverse transient occurs.
An analysis of the waveshapes of the transition shows
that if a pulse representing the difference between eO and
ee delayed is added in proper polarity of the summed signal,
that the steepness of both transitions can be restored and
the "blurring" eliminated. This is achieved by the switch-
ing inverter 52, which forms the high frequency luminance
branch (YHF). A differential amplifier, which forms the
input element of 52, produces a pulse which is equal to
the signal difference at each transition. The "difference"
pulse is coupled to a controlled phase inverter forming a
second element within 52. The phase is controlled by the
switching waveform (eSw)at twice the horizontal clock rate.
If the edge of the transition occurs between unpaired line
elements in the row, the phase inverter produces an in-
verted signal. If the edge of the transition occurs
between paired line elements in a row, the phase inverter
produces a non-inverted output. The correction pulse,
(i.e., the high frequency luminance information) is combined
in the luminance adder 53.
- 20 -
35-DV-123
~3~76~a
A first result of the double combination of the even
and delayed odd signals, in which both sum and differences
are used, is to preserve the full video spectrum without
high frequency attenuation. The second result of the
combination is a 3 dB improvement.in signal to noise
ratio.
The operation of the two correction network will now be
explained. The blue correction network is designed to
correct a false blue signal which arises in the blue channel
on the transients of a non-blue signal. The false blue
correction network consists of a two sensor element delay
45 (320 nanoseconds) and a differential amplifier 46. The
first input terminal of the differential amplifier 46 is
coupled to the output of the luminance adder 51 in the low
frequency branch. The second input terminal of the differen-
tial amplifier 46 is coupled through the two sensor element
delay 45 to the output of the luminance adder (51). The
differential output from 46 is coupled to the input adder
15, whose output is coupled with the input of the absolute
value amplifier 43. The adder 55 is typically a resistance
~ ~e s~
matrix. The ~es~r of the addition is to cancel a tran-
sient which appears in the blue output at both leading and
trailing edges of a non-blue signal.
How the false blue transient arises in the blue channel
and its correction is explained with reference to Figure
2D. Assuming similar conditions as in Figure 2C, except
that the scene is now green, the odd outputs (eO) at pre-
amplifier 31 are all "l's" and the even outputs (ee) at
preamplifier 32 are all "l's". The summed waveform ~eO+ee)
at the output of adder 40 is all "2's". When the summed
waveform is delayed and subtracted from the delayed summed
waveform at the output of differential amplifier 42, the
35~DV-123
1~3~764
non-blue (i.e., green) produces a generally "O" output
in the blue channel except for a -2 pulse at the beginning
and a +2 pulse at the end of the color bar. This is the
false blue signal which requires correction. The function
of the blue correction circuit 45 and 46 is to remove these
transients by adding an equal and opposite signal at the
input to the absolute blue amplifier 43. Neglecting the
delay in element 49, the ee waveform is applied to one
input of luminance adder 51, and the eO waveform is applied
to the other input of adder 51. When the two are added,
a double amplitude waveform (eO + ee) is produced. If this
waveform is delayed in 45 by two elements, the waveform
(eO + ee) d is formed. If the delayed waveform is subtracted
from the undelayed waveform, the blue correction is generated.
A comparision between the false blue signal and the blue
correction signal, shows that if the two are added, the sum
produces a corrected signal. The correction network is
not substantially affected by the delay in element 49 and
does produce an approximate cancellation of the non-blue
transients. The correction is normally optimized by
adjustment of the magnitude of the correction waveform.
The second correction is introduced into the luminance
channel to remove blue stripe interference. This in-
terference occurs at the same frequency as the blue stripe
in the filter layer 17. This luminance correction is
provided by the adders 48 and 50, which each have one input
coupled into the output of the blue channel differential
amplifier 42. By this connection, a signal is derived
from the blue channel, and added to the odd and even
preamplifier outputs in the adders 48 and 50 at the input of
luminance channel.
The nature and correction of the blue stripe inter-
- 22 -
-
35-DV-123
11~1769t
ference may be explained with reference to Figure 2E. In
Figure 2E, a white bar is shown focused through the filter
upon the sensor elements 7 and 18. The white signal in
the luminance channel experiences an attenuation at sensor
elements 7, 8, 11, 12 and 15, 16 due to the periodic blue
attenuation of the blue absorption stripe in filter layer
17. Assuming no correction, the white signal will produce
a luminance signal (Y uncorrected) containing periodic
fluctations at the spatial frequency of the blue filter
(as illustrated). The correction is provided by deriving
a blue signal from the blue channel in its a.c. form and
prior to rectification in the absolute value amplifier.
The decoded blue waveform, which has the desired periodicity,
is then taken in appropriate phase (normally inverted) and
scaled to form a blue correction waveform at the input
adders to 48, 50. When added to the separate odd and even
channel signals, the luminance signal is corrected as
illustrated. The correction is accurate in non-transient
portions of the white signal.
A practical circuit diagram of the color processor
is shown in Figures 3A and 3B. The design is a solid
state design using both discrete semiconductors and in-
tegrated circuits. The circuit diagram generally contains
legends in parenthesis indicating the block in Figure 1 in
which the circuit elements are formed. In particular, in
the green channel, the sample and hold circuit (block 36,
Figure 1) includes the integrated circuit (~9) and the FET
Q16. The buffer 37 in Figure 1 includes the FET Q17. In
the red channel, the differential amplifier 33 in Figure 1
is the integrated circuit ~3. In the blue channel, the
input adder 40 of Figure 1 is formed by the resistances
R78 and R79 in conjunction with Q13; the delay line 41 of
- 23 -
35-DV-123
~1~17~
Figure 1 is DL4; the differential amplifier 42 of Figure 1
is the integrated circuit ~5; the adder 55 of Figure 1 are
the resistors R88, R89, R90 and Q14; the A.B amplifier
43 of Figure 1 are the integrated circuits ~7 and ~8. In
the luminance channel, the adder 48 of Figure 1 is the re-
sistor R7 and the associated adjustable resistor and tran-
sistor Q2; the adder 50 are the resistors R4 and R5 and
transistor Ql; the delay line 49 of Figure 1 is the delay
line DL-l; the sampler 52 of Figure 1 is the integrated
circuit ~1 and FETs Q4 and Q5; the adder 51 of Figure 1 are
the resistors R31, R32, R120 and transistor Q8; the adder
53 of Figure 1 are the resistors R34 and R35 and transistor
Q9.
The present arrangement represents a near optimum
utilization of the resolution available in a given solid
state imager. Neither the filter nor the color processor
reduce that resolution. The use of two band rejection
filters (as opposed to three band pass filters) set at the
narrow bands of red and blue is consistent with a wide
band treatment in the imager output of green or luminance.
The use of two spatial frequencies; the higher for the red
rejection filter and the lower for the blue rejection filter
and the lower for the blue rejection filter, provides both
the ease in separation of the electrical signals noted
-f~, e ~ \ ~
eali~Er, and gives an optimum distribution of the available
information. One needs less red, and even less blue re-
solution to achieve a color display currently thought to
be optimum. The green is unstopped in all sections of
the color filter, and the same is approximately true in
respect to the luminance. In short, the color filters
preserve the available resolution. The color processor
also is designed to preserve both the chromatic content
- 24 -
35-DV-123
1~317~91
and the high resolution luminance information. This is
achieved by the use of separate narrow band primary color
channels which are supplemented by the addition of luminance
information of higher bandwidth.
In the illustrated embodiment, the horizontal scanner
selects sensors two at a time at a typically 320 nano-
second interval. Late on, the color processor combines the
information from the odd and even sensors in each pair and
allocates a 160 nanosecond interval to each sensor in the
consolidated electrical output signal. The delay lines
41 and 45 have a 320 nanosecond delay corresponding to the
horizontal scanning interval or twice the period allocated
to each horizontal sensor element. The delay line 49 has a
160 nanoseeond delay eorresponding to one half the horizontal
3~
f-)~' seanning interval or the prio~ alloeated to eaeh horizontal
sensor element.
While the inventive embodiment has been shown using a
eharge injection imaging device having simultaneously scanned
paired outputs, it should be evident that the inventive
principles are equally applicable to other kinds of imaging
devices. For instance, these prineiples are also applicable
to charge injeetion deviees in whieh the seanning is done
in an element by element sequential fashion or to charge
eoupled deviees.
- 25 -
