JPS641996B2 - - Google Patents

Description

[Detailed description of the invention] The present invention relates to a video signal band compression device.

When transmitting information, information degradation often occurs during the transmission process and signal processing process. One way to avoid this information deterioration as much as possible is to digitize the information source.

On the other hand, although the quality of the information can be maintained by digitizing the information source, the bandwidth occupied by the information source increases significantly. In other words, a transmission line with a large capacity is required.

This phenomenon is the same when transmitting video signals, and when considering digitizing video signals and converting them into a signal source that is less susceptible to deterioration for transmission, it is difficult to reduce the bandwidth occupied by the digitized information. The big challenge is how to narrow it down.

FIG. 1 shows a block diagram of a configuration for digitizing and transmitting a video signal. In the figure, an analog video signal is applied to a sampler 2 via a video signal input terminal 1, and after being sampled, is applied to an analog-to-digital converter 3, where it is digitized and supplied to a compressor 4. . A compressor 4 deletes unnecessary data and performs band compression, a modulator 5 performs digital modulation, and information is transmitted through a transmission path 6. Transmission line 6
The information output from the demodulator 7 undergoes the inverse operation of the modulator 5 and is applied to the expander 8. Compressor 4 with expander 8
The reverse operation is performed, and the video signal is restored to analog form by the digital-to-analog converter 9 and then sent out via the video signal output terminal 10.

In the end, the analog-to-digital converter 3 digitizes the video signal, and the compressor 4 deletes unnecessary information, thereby reducing the occupied bandwidth.

Next, a conventional example of band compression will be briefly explained below.

When the information source is a video signal, there are broadly two methods for band compression. One of them is a method of deleting redundant data included in a video signal, and the other is a method of using human visual characteristics to delete unnecessary data based on visual characteristics. Most of the practical methods belong to the former category, such as differential predictive coding (DPCM) and orthogonal transformation.

Now, let us take the case of differential predictive coding (hereinafter referred to as "DPCM") as an example.

Normally, the sampling frequency is set to three times the frequency of the color subcarrier in consideration of Nyquist's theorem and measures against beat interference with the color subcarrier. On the other hand, 8 bits are allocated to each sample in order to suppress the quantization noise that inevitably occurs during digitization to a permissible value or less. At this time, the data rate of the information source is approximately
It becomes 86Mbit/sec. (color subcarrier frequency is 3.58M
Hz, so 3.58 [MHz] x 3 x 8 [bit] = 86 [Mbit/sec]). Since the occupied bandwidth of an analog video signal is approximately 4 MHz, the occupied bandwidth of the information source increases by more than 20 times due to digitization. Therefore, the DPCM method is used to remove redundancy and reduce the occupied bandwidth.

As already mentioned, DPCM is a method for reducing the redundancy of video signals, and its principle is as follows.
It takes advantage of the strong correlation between adjacent pixels. This point will be briefly explained using the pixel diagram shown in FIG. In Figure 2, (2l-1), (2l) and (2l
+1) are (2l-1)th, (2l)th and (2l
+1)th horizontal scanning line. Each "0" mark indicates a pixel, and the broken line indicates a signal waveform. Normally, both the luminance component and the color component are almost the same within a minute area of the screen, so the carrier color signal has a waveform whose phase is inverted every horizontal scan.

Therefore, focusing on pixel x, this pixel x is very similar to pixel a, pixel b, pixel c, etc. Therefore, for example, pixel x is predicted using pixel b and pixel c to calculate the predicted value x^. That is, the predicted value x^ = 1/2 (b + c) ... (101) should be very close to the original pixel x. The difference ε between this predicted value x^ and the actual pixel x, that is, the prediction error ε=x−x^ (102) is calculated, and only the value corresponding to this error ε is transmitted. After all, since the prediction error ε is close to zero in most cases, the amount of information to be transmitted is reduced, resulting in band compression.

There has been a lot of research into this kind of DPCM, but there are many reports that it can reduce the amount of data by about half.

On the other hand, the data rate to be transmitted is greatly influenced by the sampling frequency of the sampler 2. for example,
If the sampling frequency is halved, the data rate output from the analog-to-digital converter 3 is also halved. That is, reducing the sampling frequency as low as possible is also an effective means for reducing the data rate.

Next, sub-Nyquist sampling, which is a sampling method that allows the sampling frequency to be set low, will be briefly explained below.

Sub-Nyquist sampling actively utilizes the properties of a video signal and performs sampling at a sampling frequency lower than the Nyquist frequency.

This sampling will be explained with reference to the pixel diagram shown in FIG. In Figure 3, (2l-1), (2l), (2l+
1) and (2l+2) indicate each horizontal scanning line as in FIG. 2, and white circles indicate pixels extracted by sub-Nyquist sampling. As can be seen from this pixel diagram, the sampling frequency is approximately twice the frequency of the color subcarrier. Strictly speaking, the sampling phase shifts by 1/4 period for each horizontal scan, so the horizontal scanning frequency
If f _H , then the sampling frequency f _s is becomes. On the other hand, when the frequency of the color subcarrier is f _sc , f _sc =(227+1/2) f _H (104), so n in equation (103) has a value around 455. If a sampling frequency f _s is set that satisfies equation (103) with such a value of n, the pixels will have the relationship shown in FIG. 3. In the sampling shown in FIG. 3, since sampling is performed at a frequency lower than the Nyquist frequency, the video signal shown by the broken line cannot be transmitted as it is. However, when the samples from two horizontal scans are superimposed, four of the color subcarriers
When sampling at twice the frequency, very similar pixels can be obtained. Specifically, pixel d is embedded as a pixel at position d' after two horizontal scans, and pixel e is embedded as a pixel at position e' after two horizontal scans. The same operation is performed for all other pixels. Since there is a very strong correlation between pixel d and the original pixel at position d', by embedding the pixel row two horizontal scans ago, four of the color subcarriers are
If the sample is multiplied, pixel columns that are extremely similar will be obtained, and a signal that is extremely similar to the video signal shown by the broken line will be transmitted.

In this way, by using sub-Nyquist sampling, the sampling frequency can be lowered to about 2/3 of the conventional sampling frequency, so the data rate sent out from the analog-to-digital converter can also be reduced to about 2/3 of the conventional sampling frequency. It will be reduced to

By the way, if the above-mentioned sub-Nyquist sampling is used, the data rate will be low at the sampling point, making it an extremely efficient method. Reducing the data rate is a more effective method. However, conversely, because sub-Nyquist sampling is used, conventional DPCM cannot be expected to improve the band compression effect very much. The reason for this will be explained below with reference to the pixel diagram in FIG.

In the case of conventional sampling, as can be seen from the pixel diagram in FIG. 2, there are pixels with a high degree of correlation near each pixel, so efficient prediction is possible. but,
In the case of sub-Nyquist sampling, as is clear from the pixel diagram in FIG. 3, there are no pixels with a high degree of correlation near each pixel. Specifically, when focusing on pixel f, for example, there is no pixel with a high degree of correlation nearby, and pixel g is the pixel most similar to pixel f based on the relationship of color subcarrier phases. However, since pixel g is far away from pixel f, the degree of correlation is low. In the end, even if pixel f is predicted using only pixel g, the prediction accuracy is poor, resulting in inefficient band compression.

As mentioned above, DPCM and sub-Nyquist sampling are effective means that greatly contribute to reducing data speed when viewed individually, but even if you try to use both together to further reduce data speed, No such effect can be expected.

In view of the above points, the present invention provides a video signal band compression device that uses sub-Nyquist sampling and DPCM in combination to effectively utilize the capabilities of both to further reduce the data speed. It is something.

The principle of the present invention will now be explained.

First, sampling will be explained with reference to the pixel diagram shown in FIG. In FIG. 4, the circles are pixels extracted by sub-Nyquist sampling used in the present invention, and the broken lines are video signal waveforms, (2l-1) to (2l+
2) respectively indicate horizontal scanning. As can be seen from the pixel diagram, the sampling used in the present invention samples the video signal at half the color subcarrier cycle, but the sampling phase shifts by half a cycle every two horizontal scans. Compare with each other. In such sampling, 4
Since the sampling phase and horizontal scanning phase match for each horizontal scan, the sampling frequency and horizontal scanning frequency are 1/
4 offset relationship. On the other hand, since the sampling period of each line is exactly half of the color subcarrier, the sampling frequency f _s satisfies equation (103). (However, n=
455) On the other hand, in the pixel diagram of FIG. 4, if a pixel column just two horizontal scans ago is inserted into each pixel column, a signal almost equal to the video signal shown by the broken line can be obtained due to the correlation of the video signals.

In this way, the sampling shown in FIG. 4 can be said to be a type of sub-Nyquist sampling, both from the sampling frequency and from the pixel diagram.

Next, we will explain the prediction method using DPCM.

In Figure 4, (2l) and (2l+2) are even-numbered horizontal scans, and (2l-1) and (2l+1)
is the odd horizontal scan. Now, consider the case of even-numbered horizontal scanning lines and focus on pixel x _e . Pixel h, pixel i, and pixel j have a very high degree of correlation with pixel x _e . Therefore, pixel x _e is pixel h,
Prediction can be made with high accuracy using pixel i, pixel j, etc. On the other hand, in the case of odd-numbered horizontal scanning, focusing on pixel x _p , pixel m is a highly correlated pixel. Therefore, pixel x _p can be predicted using pixel m. That is, two-dimensional prediction is performed for even-numbered horizontal scans, and one-dimensional prediction is performed for odd-numbered horizontal scans.

The above describes the sampling procedure and prediction method of the present invention,
This is a simple explanation of the principle of the present invention. As already mentioned, if sub-sub-Nyquist sampling was used, efficient prediction was not possible, but by adopting sub-Nyquist sampling shown in Figure 4, two-dimensional prediction and first-order prediction are performed for each horizontal scan. By alternately selecting the original prediction, it is possible to reduce the data rate even more efficiently with DPCM in addition to reducing the data rate with sub-Nyquist sampling.

In FIG. 5, 18 is a video signal input terminal;
9 is an analog-to-digital converter, 20 is a sub-Nyquist converter, 21 is a predictive differential encoder, 22 is a combiner, 23 is a modulator, 24 is a synchronization signal and burst signal separator, 25 is a phase synchronized oscillator, 26 is a controller, 27 is an identification signal generator, 28 is a transmission line,
29 is a demodulator, 30 is an expander, 31 is an interpolator, 3
2 is a digital-to-analog converter, 33 is a video signal output terminal, 34 is a phase synchronized oscillator, 35 is an identification signal separator, and 36 is a controller.

An analog video signal is applied to an analog-to-digital converter 19 and a synchronization signal and burst signal separator 24 via a video signal input terminal 18, and the synchronization signal and burst signal are extracted by the synchronization signal and burst signal separator 24. The extracted synchronization signal and burst signal are applied to the phase synchronization oscillator 25 and the controller 26. The phase synchronized oscillator 25 generates a continuous wave that is phase synchronized with the applied burst signal and whose frequency is multiplied. On the other hand, the controller 26 generates control signals for controlling the sub-Nyquist converter 20, the compressor 21, and the identification signal generator 27.
Further, the analog video signal input to the analog/digital converter 19 is converted into a digital signal at the timing of the clock signal generated by the phase synchronized oscillator 25. This clock signal is phase synchronized with the color subcarrier signal and has exactly four times the frequency. After all, within one cycle period of the color subcarrier, there are exactly 4
One pixel is converted into a digital value and supplied to a sub-Nyquist converter 20. In the sub-Nyquist converter 20, predetermined pixels are thinned out according to the control signal applied from the controller 26, converted to sub-Nyquist, band-compressed by the compressor 21, and then sent to the synthesizer 2.
I am guided by 2. In the synthesizer 22, the identification signal from the identification signal generator 27, which operates according to the control signal from the controller 26, and the data string of sub-Nyquist pixels are combined, and after being modulated by the modulator 23, the signal is transmitted to the transmission line 28. via phase-locked oscillator 34 and demodulator 2
It is added to 9. The phase synchronized oscillator 34 generates a continuous signal that is synchronized in phase with the signal output from the transmission line 28. On the other hand, the demodulator 29 has a phase synchronized oscillator 3.
The clock signal generated in step 4 and the signal output from the transmission line 28 are applied, and the signal supplied from the transmission line 28 is demodulated. In the demodulator 29, the modulator 23
The reverse operation is performed to reproduce the same data as the data input to the modulator 23 as the output data of the demodulator 29. The data string sent from the demodulator 29 is applied to an identification signal separator 35 to extract only the identification signal, and the operation of the controller 36 is synchronized by this identification signal. On the other hand, in the expander 30, the demodulator 2
The output data string of 9 is subjected to the inverse operation of the compressor 21 in a predetermined manner. The interpolator 31 interpolates the pixels deleted by the sub-Nyquist converter 20 with the pixels created by the expander 30, converts them into analogs by the digital-to-analog converter 32, and sends out an analog video signal via the video signal output terminal 33. do.

The outline of one embodiment of the present invention has been explained above, but the sub-Nyquist converter 20 and compressor 21 will be explained in more detail.

First, the sub-Nyquist converter 20 will be explained. As already described, the present invention uses a method of transmitting pixels as shown in the pixel diagram of FIG. That is, two pixels after sub-Nyquist conversion exist in one cycle of the color subcarrier, and the pixel position is shifted by 1/2 pitch every two horizontal scans. An example of the configuration of the sub-Nyquist converter 20 that performs this operation is shown in a circuit diagram in FIG.

In FIG. 6, 37 is an input terminal, 38 is a latch, 40 is a clock signal input terminal, 41 is a control signal input terminal, 42 and 43 are logic inverters, 44 and 45 are AND circuits, 46 is a logical sum circuit (OR). Fifth
The output data of analog-to-digital converter 19 in the figure is applied to latch 38 via input terminal 37. On the other hand, the phase synchronized oscillator 25 in FIG.
The clock signal generated by the controller 26 and the control signal generated by the controller 26 are respectively input to the clock signal input terminal 40.
and is supplied to the control signal input terminal 41. The clock signal supplied to the clock signal input terminal 40 is applied to the AND circuit 45 and at the same time is also applied to the AND circuit 42 via the logic NOT circuit 42.
Further, the control signal supplied to the control signal input terminal 41 is also applied to the AND circuit 45 and, at the same time, to the AND circuit 44 via the logic NOT circuit 43. The AND circuits 44 and 45 logically AND the two input signals, and the OR circuit 46 logically adds the outputs of both AND circuits. OR circuit 46
is applied to latch 38 as the latch 38 clock signal. In the latch 38, the OR circuit 46
The input data is stored in accordance with the pulse input from the input terminal 39, and the status is continuously outputted to the output terminal 39. The signal input to the control signal input terminal 41 is
The level is inverted every 2 horizontal scans, and 2
The high level is maintained during the horizontal scanning period, and the low level is maintained during the next two horizontal scanning periods. On the other hand, the clock signal applied to the clock signal input terminal 40 has a repetition frequency four times the color subcarrier frequency. In the end, the compressor 2
A clock signal with a timing corresponding to the pixel to be supplied to the pixel 1 is created. In this way, latch 3
8 is applied via output terminal 39 to compressor 21 in FIG.

FIG. 4 shows a pixel diagram based on sub-Nyquist sampling used in this embodiment, and a timing diagram is also shown corresponding to the pixel diagram. Timing 11 indicates the clock timing applied to the clock signal input terminal 40, and timing 12 and timing 13 indicate the timing of the pulses sent from the OR circuit 46 during the (2l-1) and (2l)th horizontal scanning periods, respectively. Timings 14 and 15 indicate the timings of pulses sent from the OR circuit 46 during the (2l+1) and (2l+2)th horizontal scanning periods, respectively. In the end, timing 12, 13, 1
4, 15... are input to the latch 38, so the output terminal 39 outputs (2l-1), (2l), (2l+1),
Data corresponding to the circled pixels described in (2l+2)... will be transmitted.

Next, an example of the compressor 21 in the embodiment shown in FIG. 5 will be explained in detail. Figure 7 shows the compressor 2
1 block diagram. In the figure, 47 is an input terminal, 48 is a subtracter, 49 is a quantizer, 50 is an output terminal, 51 is an inverse quantizer, and 52
is an adder, 56 is a first predictor, 57 is a second predictor, 55 is a switch, 53 is a control signal input terminal, and 54 is a switch controller. The output data of the sub-Nyquist converter 20 in FIG. 5 is applied to a subtracter 48 via an input terminal 47. Subtractor 48
The data added to the subtracted data is subtracted by the predicted value selected by the switch 55, quantized to a predetermined value by the quantizer 49, and supplied to the synthesizer 22 in FIG. 5 via the output terminal 50. On the other hand, the output data of the quantizer 49 is dequantized by the dequantizer 51.
Note that the term "inverse quantization" used here has the following meaning. That is, quantizer 4
9, the difference value calculated by the subtractor 48 is quantized and that value is further converted to another representative value, whereas the original quantized value is restored from this converted other representative value. It means to do. Therefore, the quantized difference value is sent from the inverse quantizer 51 to the adder 52. At the same time, switch 55
The output of , that is, the predicted value, is also supplied to the adder 52 , and the two are added together and supplied to the first predictor 56 and the second predictor 57 . The output values of both predictors are applied to a switch 55, and one of the predicted values is led to an adder 52 and a subtracter 48 via the switch 55. The control signal from the control signal generator 26 in FIG.
3 to the switch controller 54.
The switch controller 54 controls the switch 55 according to this control signal, and calculates a predetermined value from the predicted value created by the first predictor 56 and the predicted value created by the second predictor 57. to select one.

The feature of the present invention is to use both sub-Nyquist sampling and DPCM at the same time by performing sub-Nyquist sampling as shown in Figure 4, and then selecting one-dimensional prediction and two-dimensional prediction in a predetermined manner. This is what made it possible. Therefore, the procedure for selecting a prediction method will be explained below.

If the first predictor 56 in FIG. 7 performs one-dimensional prediction, the second predictor 57 performs two-dimensional prediction. Assuming that sub-Nyquist sampling is performed in the manner shown in the pixel diagram shown in FIG. 4, plane prediction is possible within even-numbered horizontal scanning periods as described above.

Usually, when predicting pixel x _e , there are many methods such as adding the increase of pixel i to pixel h to pixel j, or using the average value of pixel i and pixel j. Therefore, if the prediction function used in the even-numbered horizontal scanning period is P _E (z), then in the former prediction method, P _E (z) = z ^-4 + (z ^-908 − z ^-912 ) ... ( 105) In the latter prediction method, P _E (z) = 1/2 (z ^-4 + z ^-908 ) ... (106). however On the other hand, two-dimensional prediction is impossible within odd-numbered horizontal scanning periods. For pixel x _p , the efficient prediction point is pixel m, so if the prediction function used in the odd-numbered horizontal scanning period is P _O (z), then P _O (z) = z ^-4 ... (108) becomes. In FIG. 7, the switch 55 is controlled so that the output of the first predictor 56 is taken out during odd-numbered horizontal scanning periods, and the output of the second predictor 57 is taken out during even-numbered horizontal scanning periods.

The above is an example of the configuration of the compressor 21. Next, an example of the configuration of the decompressor 30 is shown in a block diagram in FIG.

In FIG. 8, 58 is an input terminal, 59 is an inverse quantizer, 60 is an adder, 61 is an output terminal, 65 is a first predictor, 66 is a second predictor, 64 is a switch, and 62 is a control A signal input terminal 63 is a switch controller. These inverse quantizer 59, adder 60, first predictor 65, second predictor 66, switch 64, and switch controller 63 are the inverse quantizer 51, adder 52, and first predictor in FIG. Predictor 5
6, corresponding to the second predictor 57, switch 55, and switch controller 54, respectively. Therefore, the operation will be briefly explained.

The output data of the demodulator 29 shown in FIG. 5 is dequantized by the dequantizer 59 via the input terminal 58, and then added to the predicted value supplied from the switch 64 in the adder 60. The added data is sent to the interpolator 31 in FIG. 5 via the output terminal 61 and is also applied to the first predictor 65 and the second predictor 66 at the same time. A control signal from the controller 36 in FIG. 5 is supplied to a switch controller 63 via a control signal input terminal 62, and controls a switch 64 at a predetermined timing. The predicted value selected by switch 64 is directed to adder 60.

As described above, an example of the configuration of the compressor and decompressor according to the present invention is shown in block diagrams in FIGS. 7 and 8, respectively.
5 and the second predictors 57 and 66, their configurations are determined by the prediction function.

Further, another example of the configuration of the compressor 21 according to the present invention is shown in a block diagram in FIG.

This configuration is completely the same as the block diagram shown in FIG. 7 except that the quantizer 49 and inverse quantizer 51 are controlled by a switch controller 54. The configuration shown in FIG. 9 changes the quantization characteristics between the even-numbered horizontal scanning periods and the odd-numbered horizontal scanning periods, and the prediction accuracy in the even-numbered horizontal scanning periods and the prediction accuracy in the odd-numbered horizontal scanning periods are different. This configuration was designed to reduce imbalance. For example, the prediction method of equation (108) has lower prediction accuracy than the prediction methods of equations (105) and (106). ), the error power of the former is about 5 dB larger than the error power of the latter, and if the number of bits after quantization of the former is set to 5 bits (the number of quantization steps is 32), the error power is almost the same as that of the latter. The calculation result was obtained.

As described above, almost the same prediction characteristics can be obtained by setting fine quantization steps in the case of one-dimensional prediction and, conversely, by setting coarse quantization steps in the case of two-dimensional prediction.

FIG. 10 shows another embodiment of the expander,
9 shows a decompressor corresponding to the compressor of FIG. 9. This block diagram shows the inverse quantizer 78.
The block diagram is similar to the block diagram of FIG. The inverse quantizer is similar to the inverse quantizer 59 in FIG.
The inverse quantization characteristic is controlled to change between even-numbered horizontal scanning periods and odd-numbered horizontal scanning periods in accordance with FIG. 9. Since this point is similar to that in FIG. 9, the explanation will be omitted. Also, since the components other than the inverse quantizer 59 are the same as those shown in FIG. 8, explanations thereof will also be omitted.

Above, explanations have been given with emphasis on the block diagram of FIG. 5 and the sub-Nyquist converter 20, compressor 21, and expander 30 therein shown as an embodiment of the present invention.

By the way, in the block diagram of FIG. 5, it has been explained that the phase synchronization oscillator 25 is controlled by a separate burst signal, but it is also possible to control it by only the phase synchronization signal or by both the burst signal and the horizontal synchronization signal.

On the other hand, although I have not provided a detailed explanation of the identification signal generator 27, in short, it is sufficient to be able to identify whether it is an even-numbered horizontal scan or an odd-numbered horizontal scan. A predetermined digital signal may be added to the synthesizer 22 so that this identification signal is inserted.

On the other hand, since the interpolator 31 adds the pixel column two horizontal scans ago with a delay of exactly two horizontal scanning periods, it is necessary to use a shift register or RAM (Random Access
Memory), etc.

In addition, the quantizer 49 and inverse quantizers 51 and 59 in FIGS. 7, 8, 9, and 10 are normally
Since it can be easily constructed using a ROM (Read Only Memory), switching of the characteristics of the quantizer 49 and inverse quantizers 51 and 59 in FIGS. 9 and 10 can be carried out very easily.

By the way, the transmission path 28 in FIG. 5 may be a cable or in the air, but of course it may be a magnetic recording/reproducing device for digital recording (digital VTR).
In this case, it is a tape head type.

As is clear from the above explanation, the present invention further samples the video signal using sub-Nyquist sampling.
Bandwidth compression using DPCM can also be applied, making it possible to reduce data very efficiently.

[Brief explanation of the drawing]

Figure 1 is a block diagram showing a conventional example of a digital transmission device for video signals, Figure 2 is a pixel diagram for explaining the conventional DPCM method, and Figure 3 is a diagram for explaining sub-Nyquist sampling. 4 is a pixel diagram and timing diagram for explaining the operating principle of the present invention. FIG. 5 is a block diagram showing an embodiment of the present invention. FIG. 6 is a diagram showing sub-Nyquist conversion in the present invention. FIG. 7 is a block diagram showing an example of the structure of a compressor according to the present invention. FIG. 8 is a block diagram showing an example of the structure of an expander according to the present invention. FIG. 9 is a block diagram showing an example of the structure of a compressor according to the present invention. Block diagram showing another configuration example of the compressor in
FIG. 0 is a block diagram showing another embodiment of the expander according to the present invention. 18...Video signal input terminal, 19...Analog-digital converter, 20...Sub-Nyquist converter, 2
1... Prediction differential encoder, 22... Combiner, 23... Modulator, 24... Synchronization signal and burst signal separator,
25... Phase synchronized oscillator, 26... Controller, 27... Identification signal generator, 28... Transmission line, 29... Demodulator, 3
0...Extender, 31...Interpolator, 32...Digital-to-analog converter, 33...Video signal output terminal, 34...
Phase synchronized oscillator, 35...Identification signal separator, 36...
controller.

Claims (1)

[Scope of Claims] 1. A band compression device that performs differential predictive coding after sub-Nyquist sampling a video signal, wherein the sub-Nyquist sampling has a sampling timing of half a pitch every two horizontal scans of the video signal. This is sub-Nyquist sampling in which the difference is shifted by 1, and the differential predictive coding uses two-dimensional prediction (or one-dimensional prediction) in even-numbered horizontal scanning periods, and one-dimensional prediction (or two-dimensional prediction) in odd-numbered horizontal scanning periods. A video signal band compression device characterized by differential predictive encoding. 2. Bandwidth compression of a video signal according to claim 1, characterized in that it is differential predictive coding in which the quantization characteristics in the case of one-dimensional prediction and the quantization characteristics in the case of two-dimensional prediction are different. Device.