 |
Claims  |
|
|
We claim:
1. A coding device, which transmits digital signals consisting of sampled
data obtained by sampling and quantizing an analogue video signal with a
number of quantization bits n after having compressed and coded the
digital signals, comprising:
group finishing means for forming one group for every N sampled data, N
being an integer greater than two;
reference sampled data setting means for setting one of the N sampled data
as a reference sampled data for each of the groups formed by said group
forming means;
difference calculating means for calculating the difference between at
least one of the remaining sampled data in each of said groups and the
reference sampled data;
compressing and coding means for coding the difference obtained by said
difference calculating means with a number of bits smaller in number than
the number of quantization bits n; and
a data selector for outputting at least the reference sampled data from
said reference sampled data setting means and coded differential data from
said compressing and coding means as time sequential digital data.
2. A coding device according to claim 1, wherein said difference
calculating means calculates the differences between all of the remaining
sampled data and the reference sampled data.
3. A coding device according to claim 1, wherein said compressing and
coding means codes all the differences coming from said difference
calculating means.
4. A coding device according to claim 1, further comprising:
predicted value generating means for generating at least two predicted
values on the basis of the reference sampled data for at least one among
the remaining sampled data;
means for selecting one of the generated predicted values for which the
difference between each of them and the sampled data corresponding to it
is smaller; and
coding means for representing the sampled data corresponding to the
selected predicted value by a constituted by a number of bits smaller in
number than the number of bits constituting the reference sampled data, on
the basis of the differential data between the selected predicted value
and the corresponding sampled data.
5. A coding device, which transmits digital signals consisting of sampled
data obtained by sampling and quantizing an analogue video signal with a
number of quantization bits n after having compressed and coded the
digital signals, comprising:
group forming means for forming one group for every N sampled data, N being
an integer greater than two;
reference sampled data setting means for setting one of the N sampled data
as a reference sampled data for each of the groups formed by said group
forming means;
predicted value forming means for forming a predicted value for at least
one of the remaining sampled data in each of said groups;
difference calculating means for calculating the difference between the
predicted data coming from said predicted value forming means and the at
least one of the remaining sampled data;
compressing and coding means for coding the difference obtained by said
difference calculating means with a number of bits smaller in number than
the number of quantization bits n; and
a data selector for outputting at least the reference sampled data from
said reference sampled data setting means and coded differential data from
said compressing and coding means as time sequential digital data.
6. A coding device according to claim 5, further comprising:
means for preventing the supplying of at least one of the sampled data to
said data selector so that it is not coded when N is an integer greater
than three.
7. A coding device according to claim 5, further comprising:
predicted value generating means for generating at least two predicted
values on the basis of the reference sampled data for at least one among
the remaining sampled data;
means for selecting one of the generated predicted values, for which the
difference between each of them and the sampled data corresponding to it
is smaller; and
coding means for representing the sampled data corresponding to the
selected predicted value by a code constituted by a number of bits smaller
in number than the number of bits constituting the reference sampled data,
on the basis of the differential data between the selected predicted value
and the corresponding sampled data.
8. A coding device according to claim 7, wherein said predicted value
generating means is so constructed that the generated predicted values are
the value of the reference sampled data closest in time or in distance to
the sampled data to be compressed and coded within the same line and the
value of the reference sampled data having the same phase of the color
subcarrier as that of the sampled data to be compressed and coded and
which are the closest in time or in distance to the sampled data.
9. A coding device according to claim 7, wherein the frequency of sampling
and quantizing the video signal is four times as great as the frequency of
the color subcarrier in the video signal, said coding device further
comprising:
means for selecting the reference sampled data for every N=3 samples; and
means for setting the value of the reference sampled data adjacent to the
remaining sampled data within the same line and the value of the reference
sampled data having the same phase of the color subcarrier as that of the
remaining sampled data and which are distant by four samples from the
remaining sampled data, as two predicted values corresponding to each of
the remaining sampled data other than the reference sampled data.
10. A coding device according to claim 7, further comprising:
means for generating a flag indicating the result of the selection of the
generated predicted values and supplying the flag to said data selector.
11. A coding device according to claim 7, wherein said predicted value
selecting means comprises:
means for transforming data obtained by compressing and coding each of the
at least two generated predicted values into data having the same number
of bits as that of the differential data on the basis of the transformed
data;
means for adding the generated predicted value to the transformed data; and
means for comparing the level of the result obtained by said adding means
with the level of the original sampled data which are to be compressed and
coded.
12. A coding device according to claim 7, wherein said predicted value
selecting means is composed of:
means for comparing the levels of predicted errors between each of the at
least two generated predicted values and the original sampled data which
are to be compressed and coded.
13. A coding device according to claim 7, wherein said predicted value
generating means comprises:
means for generating predicted values based on more than one of the
reference sampled data, which are close to the sampled data to be
compressed and coded, within the horizontal scanning line including the
sampled data to be compressed and coded, and M lines before and behind the
horizontal scanning line, M being an integer greater than two; and
means for generating a predicted value on the basis of more than one of the
reference sampled data having the same phase of the subcarrier as that of
the sampled data to be compressed and coded, within the horizontal
scanning line including the sampled data to be compressed and coded, and M
lines before and behind the horizontal scanning line.
14. A coding device according to claim 7, wherein said predicted value
generating means comprises:
means for generating predicted values based on more than one of the
reference sampled data close to the sampled data to be compressed and
coded, within the same field as that including the sampled data to be
compressed and coded, and M fields before and behind the field, M being an
integer greater than two; and
means for generating a predicted value on the basis of more than one of the
reference sampled data having the same phase of the subcarrier as that of
the sampled data to be compressed and coded, within the same field as that
including the sampled data to be compressed and coded, and M fields before
and behind the field.
15. A coding device according to claim 7, wherein said predicted value
generating means is composed of:
means for transforming compressed and coded data into data having a number
of bits, which is equal to that of the differential data, on the basis of
the compressed and coded data as the at least two generated predicted
values;
means for adding the generated predicted values to the transformed
differential data having the number of bits n;
means for calculating an interpolated value corresponding to the non-coded
sampled data;
means for calculating the difference between the result of said adding
means and the original sampled data to be compressed and coded; and
means for calculating the difference between the original sampled data of
the non-coded sampled data and the interpolated value obtained by each of
the generated predicted values.
16. A coding device according to claim 7, wherein the frequency for
sampling and quantizing the video signal is four times as great as the
frequency of the color subcarrier in the video signal; said device further
comprising:
means for coding one of N=3 sampled with a number of bits n as the
reference sampled data, one of the sampled data adjacent to the reference
sampled data on one side being not coded, and the other sampled data
adjacent to the reference sampled data being compressed and coded on the
basis of the difference between the generated predicted value and the
original sampled data; and
said predicted value generating means includes:
means for adopting the value of the reference sampled data adjacent to the
sampled data to be compressed and coded and within the same line, and the
value of the reference sampled data distant by four samples from the
sampled data to be compressed and coded within the same line, and has the
same phase of color subcarrier as that of the sampled data to be
compressed and coded.
17. A coding device according to claim 1, further comprising:
means for preventing the supplying of at least one of the sampled data to
said data selector as that it is not coded when N is an integer greater
than three.
18. A coding device according to claim 6 to 17, wherein the analogue video
signal is a composite video signal and the N sampled are pixels brought
together in a group for every color subcarrier period among the pixels
successively coded.
19. A coding device according to claim 6 to 17, further comprising:
calculating means for obtaining an average value of the coded sampled data,
which are adjacent or adjacent but one to the non-coded sampled data on
both the sides, the non-coded sampled data being obtained by interpolation
based on the average value.
20. A coding device according to claim 6 to 17, further comprising:
calculating means for obtaining a plurality of interpolated values based on
two coded sampled data, which are adjacent and adjacent but one,
respectively, to the non-coded sampled data on both the sides as an
optimum interpolation method, the optimum interpolation method being an
interpolation method for obtaining an interpolation value at which an
error between an original sampled value, which is not coded, and the
interpolation value is a minimum; and
means for selecting the optimum value between the two interpolated values
obtained when said calculating means generates a flag indicating the
selected interpolated value.
21. A coding device according to claim 6 or 17, further comprising:
means for locating the sampled data among the N sampled data so as to be
adjacent to the reference sampled data.
22. A coding device according to claim 6 or 17, further comprising:
means for setting the N sampled data so that the sampled data adjacent to
the sampled data, which are not coded, on both sides are either the
reference sampled data or compressed and coded data in a series of data,
in which sets of the N sampled data are arranged successively.
23. A coding device according to claim 6 or 17, further comprising:
means for setting the N sampled data so that at least one of the sampled
data adjacent to the sampled data, which are compressed and coded, on both
sides is either the reference sampled data or compressed and coded data in
a series of data, in which sets of the N sampled data are arranged
successively.
24. A coding device according to claim 6, further comprising:
means for outputting the sampled data after thinning out thereof so that
coded sampled data to be transmitted or recorded in different image signal
lines for every M image signal lines combined according to a predetermined
rule have numbers from 1 to N, which are different for different lines,
where M is an integer greater than one.
25. A coding device according to claim 6, further comprising:
means for assigning an increasing number of coding bits for compressing and
coding the difference between different sampled data among N sampled data
combined according to a predetermined rule and the predicted value
obtained on the basis of the reference sampled data with increasing
distance from the reference sampled data.
26. A coding device according to claim 6, further comprising:
means for generating a flag indicating an optimum interpolation method for
the sampled data which are not coded, and supplying it to said data
selector, the optimum interpolation method being an interpolation method
for obtaining an interpolation value at which an error between an original
sampled value, which is not coded, and the interpolation value is a
minimum.
27. A coding device according to claim 6, further comprising:
calculating means for assigning an increasing number of coding bit for
compressing and coding the difference between different sampled data among
N sampled data and the predicted value obtained on the basis of the
reference sampled data with decreasing distance from the non-coded sample
data.
28. A coding device according to claim 24, wherein the predetermined M
image signal lines are image signal which are adjacent to each other in
the same image field.
29. A coding device according to claim 24, wherein the predetermined M
image signal lines are image lines which are adjacent but one in same
image field.
30. A coding device according to claim 24, wherein the predetermined M
image signal lines are image signal lines which are adjacent to each other
in two image fields which are adjacent to each other.
31. A coding device according to claim 24, wherein the predetermined M
image signal lines are image signal lines which are the same lines in two
image fields which are adjacent but one.
32. A coding device according to claim 24 wherein the predetermined M image
signal lines are image signal lines which are the same lines in two image
fields which are adjacent to each other.
33. A coding device according to claim 32, wherein the position of the
pixel, where successive sampling for every image signal line is started,
is different for two fields which are adjacent to each other.
34. A coding device according to claim 29, wherein the position of the
pixel, where successive sampling for every image signal line is started,
is different for two lines which are adjacent to each other in a same
field.
35. A coding device according to claim 29, wherein the position of the
pixel, where successive sampling for every image signal line is started,
is different for two fields which are adjacent to each other and for two
lines which are adjacent to each other in the same field.
36. A coding device according to claim 24, wherein the number of image
signal lines M is 2. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND OF THE INVENTION
This invention relates to a coding device for transmitting digital signals
(time-sequential sampled data) obtained by sampling and quantizing
analogue video signals in the form of coded signals after having reduced
the number of quantized bits per group of sample data.
In the case where analogue video signals (image signals) are transmitted as
digital signals which have been obtained by sampling and quantizing the
analogue signals, usually seven or eight bits are believed necessary for
linear quantization as the number of quantization bits per group of sample
data (which may be called below also "image data"). When the image signals
are digitized by this linear quantization, a transmission rate of about
100 Mbps of digital signals is necessary for signals according to the
standard television system, and as to signals according to a high quality
television system which has been proposed, a transmission rate which is
more than twice as high as that stated above is required.
In a device for recording/reproducing magnetically the transmitted image,
signals described above in the form of digital signals (hereinbelow called
digital VTR), since the transmission rate is extremely high, the recording
density on a tape is substantially lowered with respect to that obtained
by a conventional VTR according to the analogue recording system, and
therefore satisfactory recording time cannot be obtained. Further, the
frequency band of the signals being dealt with is very wide, and the
working speed of the digital signal processing circuit also presents a
problem, giving rise to technical difficulties and a serious obstacle to
wide spread acceptance of this digital VTR for home use, etc.
In order to resolve this problem, heretofore a so-called high efficiency
coding method (by which image data to be transmitted are reduced by coding
them so as to lower the transmission rate) has been studied. An example
thereof is described in detail in an article "Processing of image digital
signals" (in Japanese) by Takahiko FUKINUKE, publishdd by Nikkan Kogyo
Shimbunsha (Daily Industrial Newspaper Publishing Co.).
As described in this literature (Chapter 9) a so-called differential pulse
code modulation (DPCM) has been proposed and is widely known as a method
for reducing the number of bits necessary per pixel data. By this method,
the value of a particular pixel at any moment is predicted o the basis of
values of pixels which have been already coded, and the necessary number
of bits is reduced by coding the difference (error) between the predicted
value and the value of the particular pixel at that moment.
According to this differential pulse code modulation, it is possible to
reduce the number of bits per pixel to about four or five, which is about
one half of that required according to the linear quantization method.
However the DPCM method described above has a problem which should be
resolved that influences of a coding error in a transmission system
propagate on other codes one after another (so-called error propagation).
Further, since the feedback formality is adopted in general for the
differential pulse code modulation, quantization noise is fed back and has
influences on following pixels, or vibratory noise called leak contour
pattern is produced, which gives rise to gradations, fluctuations, etc. of
the image contour portion, deteriorates extremely the image quality and so
forth. Particularly, for devices in which a high image quality is
required, it was difficult to adopt the conventional DPCM method as
described above and to put it to practical use.
SUMMARY OF THE INVENTION
The object of this invention is, in view of the prior art techniques
described above, to provide a coding device capable of suppressing signal
deterioration (error propagation, etc.) accompanying the coding to a
minimum and in addition capable of reducing the necessary average number
of bits per sampled data.
According to this invention, in order to achieve the above object, the
following measures are taken.
N digital signals (time-sequentially sampled data) obtained by sampling and
quantizing analogue video signals are brought together in one group (N
being an integer greater than 2) and coded. Among these N sampled data, at
least one, which serves as a reference, is coded with a number of
quantization bits n, which is so great that errors due to the quantization
can be neglected, and for the other sampled data the difference between
the reference sampled data and each of them is compressed and coded with a
number of bits, which is smaller than n. In this way, even if a coding
error is produced, influences thereof are restricted within the group,
where it is produced, and they are not exerted on the other groups.
Consequently error propagation doesn't occur.
According to one mode of implementing this invention, in a coding device,
in which each of the groups obtained by dividing sampled data so that each
group consists of N sampled data, is compressed and coded, image signals
are transmitted while thinning out a part of the N sampled data stated
above for every line so that they ar not overlapped on each other among M
lines of the image signals having a relatively great image correlation
within the same field, between different fields, between different frames,
etc., and all the N sampled data are apparently transmitted by
interpolating the thinned sampled data among these M lines by using the
image correlation.
Or, among N sampled data (N is an integer greater than 3) of the video
signals, at least one serves as the reference sampled data having n bits;
at least one is not coded (i.e. not transmitted); and the remaining
sampled data are compressed and coded by coding the difference between the
sampled data and the predicted value based on the reference sampled data
with a number of bits, which is smaller than n. In this way decoding is
achieved also by calculating the sampled data, which are not coded, by
interpolation, using the other coded and transmitted sampled data.
Furthermore, these two processes can be combined, so that among N sampled
data (N is an integer greater than 3) of the video signals, at least one
serves as the reference sampled data having n bits; at least one is not
coded; and for the remaining sampled data the difference between each of
the sampled data and the predicted value based on the reference sample
data is compressed and coded with a number of bits, which is smaller than
n. After that, a part of the N sampled data thus coded, which are not
overlapped on each other among M lines of the image signals having a
relatively great image correlation within the same field, between
different fields, between different frames, etc., are transmitted for
every line and interpolated by using the image correlation, and the
sampled data which are not coded are interpolated by calculation on the
basis of the remaining coded sampled data. In this way the number of bits
is significantly reduced.
Further coding, by which decoding errors are small with a small number of
bits and deterioration in the image quality is slight, can be achieved,
when the optimum interpolation calculation method producing small errors
is previously obtained by a method, by which errors between a plurality of
kinds of interpolated values and the reference sampled data are obtained
at the coding for the sampled data, which are not coded, etc. and a flag
signal indicating them is transmitted.
Next, among N sampled data one is selected as the reference data, and for
each of the remaining sampled data a plurality of predicted values are
formed. Thereafter, one of them is selected. The remaining sampled data
are coded on the basis of differential data between the selected predicted
value and the others and compressed into data having a number of bits m,
which is smaller than the number of bits n of the reference data
previously described, which are transmitted (or recorded).
In the case of the television signal of NTSC system 2, predicted values are
formed for each of the remaining sampled data. Hereinbelow, the reason
therefor will be explained. When the frequency of the color subcarrier is
designated by f.sub.sc, a method may be utilized by which the sampling is
effected with a sampling frequency of 4 f.sub.sc ; the value of the pixel
preceding a relevant pixel by 4 pixels, which are in phase in the color
subcarrier, is adopted as a predicted value (first predicted value); and
the value of the adjacent pixel is adopted as another predicted value
(second predicted value). Within the same image pattern the correlation is
higher when the first predicted value is adopted than when the second
predicted value is adopted, because the first is in phase with the color
subcarrier. Consequently, the differential data therefor are small, and
thus the former is more advantageous from the point of view that the
necessary number of bits can be reduced by coding.
However, in the case where the pixel preceding the relevant pixel by 4
pixels exists beyond an edge portion of the image pattern, since the first
predicted value has no correlation, in such a case it is more advantageous
to adopt the second predicted value, which is the value of the adjacent
pixel.
In this way two predicted values are formed for each of the remaining
sampled data and the value which is more advantageous between them is
adopted. This is the reason why two predicted value are formed. The value
which is more advantageous can be known by forming differential data for
every predicted value and comparing them. That is, it is sufficient to
adopt the predicted value, whose differential data are smaller.
Although the outline of this invention has been explained in the above,
this invention will be explained again below by using another expression.
That is, in order to achieve the above object, according to this
invention, at least one sample is coded with a number of quantization bits
n, which is so great that quantization errors can be neglected, for every
N (N is an integer greater than 2) samples of the video signals to be
transmitted, and transmitted or recorded, the sample being the reference.
For the remaining samples a plurality of predicted values, each of which
corresponds to each of the remaining samples, are calculated on the basis
of the reference sample. The remaining samples are transformed into data
of a number of bits m, which is smaller than the number of bits n
described above, on the basis of differential data between each of the
predicted values and the reference sample.
The data of the number of bits m, obtained by using the predicted values,
are extended and transformed into data of a number of bits, which is equal
to that of the differential data, on the basis of the data by means of
transforming means equivalent to that used at the coding, and each of the
predicted values stated above corresponding to each of the extended and
transformed data is added thereto. That is, at the coding, provisional
decoding is effected. Next a predicted value is selected, which can give
provisionally decoded data, for which the level differences between each
of the added data of the number of bits n, which is the provisionally
decoded data, and the original sample data are the smallest, by comparing
their levels.
On the other hand, for the remaining samples the difference between the
selected predicted value and each sample is compressed and coded.
According to this invention no accumulation of quantization errors due to
the differential pulse code modulation occurs owing to the fact that for N
samples a reference sample and compressed samples, for which the
difference between the reference sample and each of the predicted values
is quantized, are coded. Further, error propagation due to coding errors
produced on a transmission path are prevented from continuing over a long
period of time, and in this way it is possible to suppress deterioration
of the image quality to the minimum.
Furthermore, according to this invention, since the coder can be
constructed in the feed forward formalism, it is possible to remove the
noise production described previously, which gave rise to a problem in the
prior feedback type coder.
On the other hand it is possible to reduce remarkably the average number of
bits per pixel by thinning out sampled data so that they are not
overlapped between different lines and transmitting the N sampled data for
every M image signal lines having a great image correlation. The thinned
out sampled data can be obtained by interpolation using other lines among
the M image signal lines. Further, it is possible to reduce the average
number of bits per pixel also by the fact that a part of the N sampled
data are neither coded nor transmitted, and it is obtained by
interpolation using a value calculated on the basis of the other coded and
transmitted sampled data among the N sample data. In this way decoding
including small errors with a small number of bits is possible, because
for this interpolated value, a plurality of interpolation calculation
methods are previously determined; interpolation errors according to each
of these methods are obtained at the coding; a flag indicating which
method gives the smallest errors is transmitted; and the optimum
interpolation method can be determined at the decoding, referring to the
flag stated above. Furthermore, a remarkable decrease in the number of
bits can be obtained by combining the thinning out of sampled data for
every M image signal lines and the method by which interpolation is
effected in the N sample data.
On the other hand, errors produced by the compression and extension can be
kept to a minimum by calculating a plurality of predicted values for the
samples to be compressed and coded; provisionally decoding them by using
each of these predicted values at the coding; selecting a predicted value
giving the provisionally decoded data, for which the level difference
between the predicted value and the original sample data is the smallest,
and compressing and coding the difference between this selected predicted
value and the reference sampled data.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is a block diagram illustrating an embodiment of the coding device
according to this invention;
FIGS. 2A and 2B show a waveform and timing for explaining operation of the
coding device indicated in FIG. 1;
FIG. 3 shows coding and decoding characteristics;
FIG. 4 is a block diagram illustrating a decoding device for the signals
coded by the coding device illustrated in FIG. 1;
FIG. 5 is a timing chart for explaining the work of the decoding device
indicated in FIG. 4;
FIG. 6 is a block diagram of an example to which this invention is applied;
FIG. 7 is a block diagram illustrating an embodiment of the data
interpolating device according to this invention;
FIGS. 8A and 8B are timing charts for explaining the operation of the
device indicated in FIG. 7;
FIG. 9 is a block diagram illustrating another embodiment of the coding
device according to this invention;
FIGS. 10A and 10B show a waveform and a timing chart, respectively, for
explaining the operation of the device indicated in FIG. 9;
FIG. 11 shows another example of coding and decoding characteristics;
FIG. 12 is a block diagram illustrating an example of the decoding device
for the signals coded by the device indicated in FIG. 9;
FIG. 13 is a timing chart for explaining the operation of the device
indicated in FIG. 12;
FIGS. 14A and 14B are timing charts for explaining still another embodiment
of the coding device according to this invention;
FIGS. 15 to 21 are diagrams illustrating examples of the structure of
various samples coded by the coding device according to this invention;
FIG. 22 is a block diagram illustrating an example of still another coding
device to which this invention is applied;
FIG. 23 is a block diagram illustrating an embodiment of the coding device
according to this invention applied to the device indicated in FIG. 22;
FIG. 24 is a timing chart for explaining the work of the device indicated
in FIG. 23;
FIGS. 25A and 25B show waveforms for explaining the principle of the
operation of the device indicated in FIG. 23;
FIG. 26 is a block diagram illustrating a specific example of a decoding
device for the coding device indicated in FIG. 23;
FIG. 27 shows the timing of signals in various parts of the circuit
indicated in FIG. 26;
FIG. 28 is a block diagram illustrating a specific example of the predicted
value selecting circuit of FIG. 23;
FIG. 29 shows the timing of signals in various parts of the circuit
indicated in FIG. 28 for the purpose of explaining the operation thereof;
FIG. 30 is a block diagram illustrating another specific example of the
predicted value selecting circuit of FIG. 23;
FIG. 31 shows waveforms for explaining the principle of operation of
another specific example of the predicted value calculating circuit;
FIG. 32 shows a waveform for explaining the principle of operation of
another embodiment of this invention; and
FIG. 33 shows waveforms for explaining the principle of operation of still
another embodiment of this invention.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Hereinbelow this invention will be explained in more detail, using
preferred embodiments. FIG. 1 is a block diagram illustrating an
embodiment of the coding device according to this invention; FIGS. 2A and
2B show a waveform and the timing, respectively, for explaining the work
thereof; FIG. 3 shows an example of coding characteristics thereof; FIG. 4
is a block diagram illustrating an embodiment of the decoding device for
the signals coded according to this invention; and FIG. 5 is a timing
chart for explaining the work thereof.
In FIG. 1 reference numeral 101 is an input terminal through which digital
image signals, which are successively sampled and are to be coded, are
inputted; 102 to 105 are delay circuits, each of which has a delay time
equal to the sampling period .tau.; 106 to 110 are data latches; 111 and
112 are subtracters; 113 and 114 are ROMs effecting the compression and
coding; 115 is an average value calculating circuit; 116 is an
interpolated value calculating device; 117 is a flag generating circuit
for generating the flag indicating the optimum interpolation method; 118
is a data selector; 119 is a control circuit for controlling the data
selector; and 120 is a coded signal output terminal.
A digital signal A.sub.i (i is an integer) of n bits, obtained by
successively quantizing the image signal V indicated in FIG. 2A, with a
period .tau. is inputted through the input terminal 101. The number of
quantization bits n is a value which is so great that errors produced by
quantization using it are negligibly small, and in this embodiment, where
image signals are dealt with, it is decided that e.g. n=7.
This invention is characterized in that the coding processing is effected
for every group of N sampled data (N being an integer greater than 2).
FIGS. 1, 2A and 2B show an embodiment in which N=4. In this embodiment,
among four sampled data represented by (A.sub.4i-1 A.sub.4i, A.sub.4i+1,
A.sub.4i+2), as indicated in FIG. 2A, the sampled data A.sub.4i indicated
by a mark o, are selected as the reference sampled data and are coded with
n bits.
Hereinbelow, these reference sampled data are represented by the same
symbol A.sub.4i. In addition, the sampled data A.sub.4i+1, indicated by a
mark x, are not coded and therefore are not transmitted. This sampled
data, which are not coded are obtained at the decoding by an interpolating
calculation using other sampled data. For the other two sampled data
A.sub.4i-1 and 4.sub.4i+2, indicated by marks .DELTA., the difference
between a predicted value obtained from the reference sample data A.sub.4i
and each of the sampled data is compressed and coded. In this embodiment
the reference sample data A.sub.4i and the difference therefrom are
obtained by using the following equations:
##EQU1##
These two differences are coded with a number of bits m (<n). In this
embodiment these two differential data B.sub.4i-1 and B.sub.4i+2 are coded
to compressed differential data C.sub.4i-1 and C.sub.4i+2 of m=4. Further,
in this embodiment, in order to reduce interpolation errors for the sample
data A.sub.4i+1 stated above, which are not coded, a flag F.sub.4i+1
indicating the optimum interpolation method is transmitted. In this
example two kinds of interpolation methods are used, and the number of
bits of F.sub.4i+1 is 1. In this way 16 bits (7+4.times.2+1) are used for
N=4 sampled data, and therefore the average number of bits per pixel is
16/4=4. Consequently the number of bits is reduced to 4/7 with respect to
that required by the method by which every pixel is coded with 7 bits.
The decrease of the number of bits based on the principle stated above is
effected as follows. A digital image signal a (a in FIG. 2B) of n bits
inputted through the input terminal 101 in FIG. 1 is inputted to the delay
circuits 102 to 105 one after another and at the same time is inputted to
the data latches 106 to 110 together with outputs of the delay circuits
102 to 105. The data are taken out from the data latches 106 to 110 with
an interval which is four times as long as the sampling period .tau., and
the data latch outputs b to f are indicated by b to f in FIG. 2B,
respectively. Among them the reference sampled data (A.sub.4i) are taken
out from the data latch 109, and the output thereof e (e is FIG. 2B) is
inputted to the data selector 118 and the two subtracters 111 an 112. The
next but one sampled data (H.sub.4i+2) to the reference sampled data
(A.sub.4i) is taken out from the data latch 107, and the output thereof c
(c in FIG. 2B is inputted in the subtracter 111. The subtracter 111 forms
the difference between the outputs c and e and a differential output g (g
in FIG. 2B) of n+1 bits is obtained. The sampled data (A4i-1) preceding
the reference sample data (A.sub.4i) are taken out from the data latch
110, and the output thereof f (f in FIG. 2B) is inputted to the subtracter
112. The subtracter 112 forms the difference between the outputs f and e,
and a differential output h (h in FIG. 2B) of n+1 bits is obtained. The
outputs g and h of these subtracters 111 and 112 are inputted in ROM 113
and ROM 114, respectively, and transformed into compressed differential
data i and j (i and j in FIG. 2B) of m (=4) bits. FIG. 3 shows an example
of the transformation characteristics of ROM 113 and ROM 114 for n=7 and
m=4.
Sixteen (i.e. equivalent to 4 bits) data, in the whole corresponding to
a.sub.0, a.sub.1 . . . a.sub.7 and b.sub.0, b.sub.1, . . . b.sub.7
indicated in FIG. 3, are written in ROM 113 and ROM 114. Among these data
those whose address is specified, corresponding to the outputs g and h of
n+1 (=8) bits from the subtracters 111 and 112, are read out. As an
example, as indicated in FIG. 3, when the value of g or h (i.e., the value
of the differential data B.sub.i) is comprised between 46 and 62, data
C.sub.i corresponding to a.sub.5 is outputted from the ROM.
On the other hand the output c of the latch 107 and the output e of the
latch 109 are inputted to the average value in FIG. 2B). The output
thereof D.sub.4i+1 is represented by:
##EQU2##
Further the output b (b in FIG. 2B) of the latch 106, the output c of the
latch 107, the output e of the latch 109, and the output f of the latch
110 are inputted to the interpolated value calculating circuit 116, which
calculates an interpolated value, which is data corresponding to the
output d (d in FIG. 2B) of the latch 108, on the basis of these values. In
this embodiment an average value l (l in FIG. 2B) of an exterpolated value
based on the outputs b and c and an exterpolated value based on the
outputs e and f is outputted. That is, the output thereof E.sub.4i+1 is
represented by:
##EQU3##
The output k of the average value calculating device 115, the output l of
the interpolated value calculating device and the output d of the latch
108 are inputted to the flag generating circuit 117. The flag generating
circuit 117 calculates an error included in the interpolated value
D.sub.4i+1 of A.sub.4i+1 obtained by the average value calculating device
115 and an error included in the interpolated value E.sub.4i+1 of
A.sub.4i+1 obtained by the interpolated value calculating device 116 and
outputs the result obtained by comparing these quantities as the flag
output m (m in FIG. 2B). That is, in this embodiment the flag output
F.sub.4i+1 is flag data of 1 bit, which is 0 when the error of D.sub.4i+1
is smaller and 1 when the error of E.sub.4i+1 is smaller.
The output e of the data latch 109, the output i of ROM 113, the output j
of ROM 114 and the output m of the flag generating circuit I17 are
inputted to the data selector 118. The data selector 118 selects them one
after another, depending on the number of bits of each of the signal
outputs, responding to instructions from the control circuit 119, and the
output thereof n (n in FIG. 2B) is outputted through the coded signal
output terminal 120. The output signal n consists of e.g. C.sub.4i-1 of 4
bits A.sub.4i of 7 bits, F.sub.4i+1 of 1 bit and C.sub.4i+2 of 4 bits in
the order indicated by n in FIG. 2B. Other pixels are dealt with in the
same way and signals therefor are outputted. Further, depending on the
structure of the system, it may be possible to modify the control signal
from the control circuit 119 and to change the order of the output signal
n of the data selector so that the signals are dealt with in units of 8
bits, e.g. 8 bits for C.sub.4i-1 and C.sub.4i+2 and 8 bits for A.sub.4i
and F.sub.4i+1. As explained above, by means of this coder it is possible
to reduce the number of bits from 7 bits for each of 4 pixels, i.e. 28
bits in total, to 16 bits, i.e. by a factor of 4/7. Here, in the output
of the data selector 118 indicated by n in FIG. 2B the position and the
length of e.g. C.sub.4i-1' A.sub.4i, F.sub.4i+2 and C.sub.4i+2 indicated
in the figure don't represent correctly the relation between different
signals in the output timing, the number of bits, etc.
Now the work of the decoding device according to this invention will be
explained, referring to FIGS. 4 and 5. In the embodiment indicated in FIG.
4, the signal coded by the coder in the embodiment explained above and
indicated in FIG. 1, whose number of bits is reduced by a factor of 4/7 in
unit of 4 pixels of 7 bits, is decoded.
In FIG. 4, reference numeral 401 is a signal input terminal, through which
the coded signal, whose number of bits per pixel is reduced, is inputted;
402 is a data distributing circuit, which distributes signal data
corresponding to each of the pixels; 403 to 406 and 412 to 415 are data
latches; 407 and 408 ar ROMs extending compressed and coded data; 409 and
410 are adders for decoding differential data; 411 is a flag decoder which
decodes the flag indicating the optimum interpolation method and outputs a
data select signal; 416 is an interpolated value calculating device; 417
is an average value calculating device; 418 is a interpolate value
selector which selects and outputs the optimum interpolated value; 419 is
a data selector for changing over data of the pixels and outputting them
one after another; and 420 is a decoded signal output terminal.
A signal indicated by o in FIG. 5, whose number of bits is reduced by the
coder, an example of which is indicated in FIG. 1, is inputted through the
signal input terminal 401. This input signal o is divided into a plurality
of data groups corresponding to the pixels by the data distributing
circuit 402, which are sent to the data latches 403 to 406, respectively.
In this embodiment, since the number of sampled data of one group dealt
with to be coded N=4, the data divided into 4 groups are inputted to the
data latches 403 to 406, respectively. The data latches 403 to 406 take
out data with a period which is 4 times as long as the sampling period
.tau., the outputs thereof p to s being indicated by p to s in FIG. 5.
Among them the data latch 405 takes out the reference sample data
(A.sub.4i) quantized with n (=7) bits by the coder and the output thereof
r (r in FIG. 5) is inputted in the data latch 414 as well as the two
adders 409 and 410. The data latch 403 takes out the compressed
differential data (C.sub.4i+2) corresponding to the sample data, which are
next but one to the reference sample data (A.sub.4i), and the output
thereof p (p in FIG. 5) is inputted in ROM 407. The data latch 406 takes
out the compressed differential data (C.sub.4i-1) corresponding to the
sample data preceding the reference sample data (A.sub.4i) and the output
thereof s (s in FIG. 5) is inputted in ROM 408. Further the data latch 404
takes out the flag data (F.sub.4i+1) indicating the optimum interpolation
method for the sample data (A.sub.4i+1), which has not been coded and
therefore has not been transmitted, and the output thereof g (g in FIG. 5)
is inputted in the flag decoder 411.
ROMs 407 and 408 extend compressed data of m (=4) bits to data of n+1 (=8)
bits according to the characteristics indicated in FIG. 3 explained above.
When the outputs p and s of the data latches 403 and 406, which are
compressed differential signals, are given thereto as addresses,
differential data (B.sub.4i+2, B.sub.4i-1) extended to 8 bits are
outputted therefrom, respectively. As an example, in the case where the
output data (C.sub.i) of the data latch 403 or 406 are a value
corresponding to a.sub.5' data (B.sub.i) having a value of 54 are
outputted as an output t or u (t or u in FIG. 5) of ROM 407 or 408. The
outputs t and u of ROMs 407 and 408 are inputted to the adder 409 and the
adder 410, respectively, where the reference sampled data (A.sub.4i),
which are the output r of the data latch 405, is added thereto. The adder
409 and the adder 410 effect operations represented by the following
equations:
##EQU4##
respectively, and in this way the sa | | |
