JPH0342556B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION A digital VTR has been proposed that converts a color video signal into a PCM signal for recording and playback.

When considering a digital VTR for broadcasting stations, in order to ensure high image quality, the video signal needs to be transmitted at a high bit rate, for example, in multichannel mode, without band compression.

On the other hand, for products that are low-priced and do not require very high image quality, such as popular models for home use, there may be a deterioration in image quality that does not cause visual problems, but the bandwidth It is necessary to compress it so that it can be transmitted at a low bit rate.

As a method to satisfy this requirement, A-D converted color video signal data is first compressed by sub-Nyquist sampling, and then subjected to differential pulse code modulation (hereinafter referred to as DPCM) to significantly compress the band and reduce the transmission bit rate. There are some things that lower the value to a desired value.

First, sub-Nyquist sampling will be explained.

That is, for example, an analog NTSC color video signal has a frequency that is, for example, four times the color subcarrier frequency f _SC .
After being sampled at 4f _SC and converted to digital PCM video data, it is resampled at a sub-Nyquist frequency f _SN (a frequency less than twice the signal band) and compressed.

This sub-Nyquist frequency f _SN is selected as follows.

In other words, in the NTSC color video signal, the frequency spectra of the color signal C and the luminance signal Y are S _C , S _Y
are in a frequency interleaving relationship as shown by the solid line in FIG. Therefore, the sub-Nyquist frequency f _SN is the color signal C shown by the broken line in FIG.
and the downward return component N _CY of the luminance signal Y is the color signal C
and the spectra S _C and S _Y of the luminance signal Y are set so that they are further frequency interleaved.

Therefore, the sub-Nyquist frequency f _SN is selected such that, for example, f _SN =2f _SC -1/4f _H (1) (where f _H is the horizontal scanning frequency).

The signal with frequency f _SN shown in equation (1) has frequency 2f _SC
can be obtained equivalently and easily by inverting the phase of the signal every two horizontal scanning synchronizations.

Therefore, in this case, what was sampled at 4f _SC is further sampled at 2f _SC , so the data is compressed to 1/2 by sub-Nyquist sampling.

The data compressed signal is further compressed by a DPCM encoder.

By the way, a DPCM encoder predicts input data, obtains the difference between the input data and this predicted data, that is, a prediction error, and transmits this error, so in order to increase the compression ratio, the prediction error must be It is desirable to predict the characteristics such that . For this purpose, it is preferable to perform two-dimensional prediction using not only the horizontal correlation but also the vertical correlation of the video signal.

However, as shown in Figure 2, in the sub-Nyquist sampled signal as described above, the sampling phase is inverted every two lines, so if the data of the previous line with a strong correlation is used as is, it will not be processed as a DPMC code. The disadvantage is that it cannot be used to predict the

That is, FIG. 2A shows sampling clock CP ₀ with frequency 4f _SC , FIG. 2 B shows sampling clock CP ₁ with frequency 2f _SC , and FIG. 2 C shows phase inverted signal ₁ of clock CP ₁ . Each clock
For example, assume that the rising time of CP ₀ , CP ₁ , ₁ is the sampling time. Figures D to G are diagrams showing the sampling phases of four lines of a certain field sampled by these clocks CP ₀ , CP ₁ , and ₁ together with the phase of the color subcarrier, and are indicated by . in the figure. 1 indicates the sampling point by clock CP ₀ , and the . surrounded by ○ indicates the sampling point resampled by clock CP ₁ and ₁ , respectively. In other words,
The lines shown in D and E of the same figure are drawn by clock _CP1 .
The lines shown in F and G in the same figure are drawn by clock CP ₂ .
This is the case where each sample is taken.

As is clear from FIGS. 2D to 2G, the sampling phase of the sub-Nyquist sampled signal is inverted line by line.

Therefore, conventionally, when DPCM-encoding this sub-Nyquist sampled signal,
Because of this discontinuity in the sampling phase, it was necessary to form a predicted signal from only information within one line. For this reason, the prediction error was relatively large, and significant data compression was not possible.

In view of the above points, the present invention aims to provide an apparatus for DPCM encoding a sub-Nyquist sampled signal, which is capable of particularly highly efficient data compression.

That is, in this invention, we focus on the fact that the sampling phase in sub-Nyquist sampling is in phase every two lines, and perform two-dimensional encoding in units of two lines that are in phase.
This is intended to improve the prediction characteristics of the DPCM encoder.

An example of this invented device is shown below as a digital VTR.
Let's explain the case using the figure as an example.

FIG. 3 shows the recording system of a digital VTR, and FIG. 4 shows its reproduction system.

In FIG. 3, reference numeral 1 denotes an input terminal for an NTSC color video signal, through which the color video signal is supplied to an AD converter 2. Also, through terminal 3, the frequency
A clock _CP0 of 4f _SC is supplied to this A/D converter 2, an input color video signal is sampled by this clock _CP0 , and the sampled value is converted into, for example, an 8-bit PCM signal.

In this case, since there is no need to transmit the horizontal blanking period, the A-D converter 2 samples only the video information portion of the horizontal scanning period TV excluding _the horizontal blanking period T _B , as shown in FIG. Ru.

In this way, one line of video data is sequentially obtained from the A/D converter 2 with intervals of time T _B (time slots).

Digital data from this A-D converter 2 is supplied to a sub-Nyquist sampling circuit 4.
In addition, the frequency of the clock CP ₀ is divided by 1/2 by the frequency divider 5 to become the clock CP ₁ of the 2f _SC , and this clock CP ₁
is supplied to the phase inversion switching circuit 6. Then, through terminal 7, a signal that inverts the state every two lines
SW ₁ is supplied to this switching circuit 6 so that clock CP ₁ and its inverted clock ₁ are alternately obtained from this switching circuit 6 every two lines.

The clocks CP ₁ and _{CP 1} are then supplied to the sub-Nyquist sampling circuit 4, and the PCM data is resampled as described above.
Data is compressed to 1/2.

The data compressed by this sub-Nyquist sampling is sent directly to the adder 9 and the subtracter 10.
The signal is also supplied to the adder 9 and subtracter 10 after being delayed by one line by the one-line memory 8. Therefore, the adder 9 and the subtracter 10 calculate the sum and difference of the data between the two lines.

As shown in Figure 2 D to G, in the NTSC color video signal, the phase of the color subcarrier is inverted for each line, so when the sum of the signals between two lines is taken, the high frequency component The color signal component C, which is , is erased, and a luminance signal Y, which is a low frequency component, is obtained as shown in FIG. Furthermore, when taking the difference between the signals between two lines, the luminance signal with a strong vertical correlation, that is, the luminance signal component Y _L with a low vertical frequency, is eliminated, and the color signal C and the vertical correlation are removed, as shown in FIG. In other words, a composite signal with the luminance signal component Y _H having a high vertical frequency is obtained.

Therefore, the adder 9 obtains a luminance signal Y, and the subtracter 10 obtains a color signal C and a luminance signal component _YH with a high vertical frequency, which are output through level adjusters 11 and 12, respectively. The signal is supplied to the first and second DPCM encoders 13 and 14.

These first and second DPCM encoders 13 and 14 include subtracters 131 and 141, predictors 132 and 142, nonlinear compressors 133 and 143, and adders 134 and 144, for example.
It consists of code converters 135 and 145 made up of ROM (read only memory), and the sum output and difference output between two lines are sent to a subtracter 1, respectively.
31 and 141.

Then, in the subtracter 131, the predictor 13
The difference between the predicted signal from 2 and the luminance signal Y is taken, while in the subtracter 141, the prediction signal from the predictor 142
The difference between the predicted signal and the color signal C etc. is taken, and a predicted error signal is obtained respectively.

In this case, the predictors 132 and 142 have characteristics suitable for each, and consideration is given to minimizing the prediction error signal.

That is, the predictor 132 for the luminance signal Y
uses the previous sample value in the prediction function. In other words, the prediction function P(Z)=Z ^-1 .

This is done in order to improve the step response for the luminance signal Y. Since the sample value just before this is information on the sum of two lines, using this value increases the vertical correlation and horizontal correlation. This is because it is used and has the strongest correlation.

On the other hand, the predictor 142 for the color signal C uses a sample value two samples before, that is, one cycle before the color subcarrier, for the prediction function. In other words, the prediction function P(Z)=Z ^-2 .

This is done because the color signal C must be predicted using sample values that are in phase with the color subcarrier. In other words, as is clear from FIGS. 2D to 2G, the immediately preceding sample value has a phase shift of π with respect to the color subcarrier, and cannot be used for prediction.

In this way, predictions suitable for each of the sum output and difference output between the two lines are made.

Prediction error signals from subtractors 131 and 141 are supplied to nonlinear compressors 133 and 143, respectively. In these nonlinear compressors 133 and 143,
By utilizing the statistical and visual characteristics of images, the prediction error signal is often small-level, and large-level prediction errors do not require much visual accuracy. The quantization level is compressed nonlinearly.

That is, the compression characteristics of these compressors 133 and 143 are as shown in FIG.
In the large level part, the number of quantization levels is nonlinearly compressed so that it becomes rough. For example, input signals of 512 quantization levels of these compressors 133 and 143 are rounded to output signals of 15 quantization levels.

The output signals of these compressors 133 and 143 are supplied to code converters 135 and 145, respectively, and the 15 quantization level signals are respectively converted into 4-bit data.

Therefore, for example, a signal that has been quantized at 8 bits is converted to a signal with 15 quantization levels by DPCM, and quantized at 4 bits by code converters 135 and 145.
It will be converted into bit digital data.

Note that the output signals of these nonlinear compressors 133 and 143 are supplied to adders 134 and 144, respectively, and added to the predicted signals from predictors 132 and 142, respectively. 14 input signals, that is, demodulated signals of the sum output and the difference output between two lines, are obtained and supplied to the predictors 132 and 142 to perform the above-mentioned prediction.

The sum output from the DPCM encoder 13 thus obtained, that is, the DPCM signal of the luminance signal Y, is supplied as is to one input terminal of the switch circuit 16. Further, the difference output from the DPCM encoder 14, that is, the DPCM signal such as the color signal C, is delayed by one line by the one line memory 15 and is supplied to the other input terminal of the switch circuit 16.

Here, for example, two lines H ₁ and H ₂ , H ₃ and H ₄ ,
Assuming that H ₅ and H ₆ ... are lines with the same sampling phase, 1 line memory 8
When the input signal is as shown in FIG. 9A, the output is as shown in FIG. 9B, so the adder 9 outputs one horizontal signal as shown in FIG. In every other horizontal interval _THA , the sum of data between two lines with the same sampling phase is obtained, and from the subtracter 10, as shown in _FIG . The difference in data between two lines is obtained. Therefore, the output of the 1-line memory 15 is E in the same figure.
As shown in , the remaining 1 adjacent to the horizontal section T _HA
Difference data between these two lines can be obtained in every horizontal interval _THB .

Then, the switch circuit 16 uses a signal SW ₂ (FIG. 9F) that inverts the state every horizontal interval from the terminal 17 to switch one input terminal, that is, the output signal selection side of the DPCM encoder 13, in the horizontal interval T _HA . In the horizontal section _THB , the signal is switched to the other input end, that is, to the output signal selection side of the DPCM encoder 14. Therefore, as shown in FIG. (including the signal _YH ) are obtained alternately for each line.

The output signal of this switch circuit 16 is supplied to an error correction encoder 18, and a parity V _PA for data error detection and correction is added to the data. In this case, as mentioned above, the blanking period T _B is used as a time slot for each line of data, so the parity data V _PA is added after one line of data D _V as shown in Figure 5. be done.

The output of the error correction encoder 18 is supplied to the recording processor 19, and as shown in FIG. 5B, the block synchronization signal BS, which is used as a reference when extracting one line of data, Added to the beginning. The data from this block synchronization signal BS to the parity V _PA is taken as one block's worth of data.

Further, in this recording processor 19, modulation as described below is performed when recording digital signals, and parallel digital signals are converted into serial digital signals.

This serial digital signal is supplied to a switch circuit 20.

This switch circuit 20 synchronizes with the rotation of the rotary head drum motor and the vertical synchronization signal of the input video signal, and changes the state in which the signal is supplied to the rotary head HA through the amplifier 22A every unit time, and the state in which the input video signal is supplied to the rotary head HA through the amplifier 22B.
The state supplied to the rotary head HB through
These heads HA and HB are alternately switched
Digital data is recorded by forming separate tracks depending on the data.

In this case, in this example two rotating heads
The structure of a VTR with HA and HB is the same as that of conventional popular VTRs, but in order to make the 1-bit width of digital data correspond to the magnetization reversal width on the tape, the tape speed and drum speed have been changed. Both rotational speeds are said to be twice that of conventional VTRs. Therefore, video PCM data for 1/2 field is recorded on one track recorded on the tape. The state is alternately switched every 1/2 field by the signal _SW3 which inverts the state.

Therefore, the recording format of the signals on the tape is similar to that of a conventional home VTR, but the heads HA and HB are in charge of the upper and lower parts of one screen separately. Further, in this case, recording is performed so that a slight guard band is formed between adjacent tracks.

In this case, the M-sequence scrambling method is used as the modulation method for recording the digital signal, so as not to increase the data rate.

Next, let's explain the reproduction system.

The signals reproduced by the rotary heads HA and HB are supplied to one and the other input ends of the switch circuit 24 through amplifiers 23A and 23B, respectively. The switch circuit 24 is switched every 1/2 field period in synchronization with the rotation of the rotary heads HA and HB by the signal _SW4 from the terminal 25, so that the signal from the rotary head HA and the signal from the rotary head HB are switched. are obtained alternately and supplied to the reproduction processor 26.

This reproduction processor 26 extracts digital data from the reproduction signal, restores the scrambled data, and further extracts digital data from the serial signal.
It is converted back into a bit-parallel digital signal. moreover,
The block synchronization signal BS is detected and the data in blocks is detected.

The output signal of this processor 26 is supplied to an error correction decoder 27, and the video data within the range that can be corrected by the parity _VPA is corrected.

The error corrected data is sent to the error correction circuit 28.
One block of data that has an error that could not be corrected, that is, one line of data, is corrected by replacing it with one line of data from the previous line, one field, or one frame. .

After this error correction, the data is delayed by one line by the one line memory 29.
The signal is supplied to the DPCM decoder 30, and is also supplied as is to the DPCM decoder 31.

The DPCM decoders 30 and 31 are composed of code converters 301 and 311 made of ROM, for example, adders 302 and 312, and predictors 303 and 313, and input data is supplied to the code converters 301 and 311, respectively. be done. Then, in code converters 301 and 311, the 4-bit input data is returned to an 8-bit difference signal with 15 quantization levels, and adders 302 and 312
are supplied respectively. And this adder 30
2 and 312, predictors 303 and 313
It is added to the predicted signal from the source and demodulated to the original 8-bit digital video data.

The demodulated digital video data is supplied to predictors 303 and 313, respectively, and is used as a signal for prediction.

In this way, the data demodulated by the DPCM decoders 30 and 31 are supplied to the adder 32 and the subtracter 33, respectively.
A subtracter 33 calculates the sum and difference between the demodulated data from the DPCM decoder 30 and the demodulated data from the DPCM decoder 31.

The sum output from the adder 32 is supplied to one input terminal of the switch circuit 35, and the difference output from the subtracter 33 is delayed by one line by the one line memory 34 and is supplied to the other input terminal of the switch circuit 35. Supplied.

Here, as shown in FIG. 10A, the data reproduced by the heads HA and HB is divided between two lines with the same sampling phase, _H1 and _H2 .
, between H ₃ and H ₄ , between H ₅ and H ₆ , etc., the sum (luminance signal Y) and the difference (chrominance signal C, including the luminance signal Y _H with high vertical frequency), etc., for each line. They are obtained alternately.

Therefore, the output data of the 1-line memory 29 is as shown in FIG.
From then on, as shown in C of the same figure, 1 of every horizontal section
In the horizontal section T _HC , every other line H ₁ , H ₃ , H ₅
The data of... is obtained, and from the subtractor 33, the data D of the same figure is obtained.
As shown in the figure, data for every other remaining line H ₂ , H ₄ , H ₆ , . . . is obtained in the same horizontal section T _HC . The output of the 1-line memory 34 is this subtracter 3
Since the output from line 3 is delayed by one _line , the data on lines H ₂ , H ₄ , _{H 6} . Become.

Then, the switch circuit 35 uses a signal SW ₅ (FIG. 10F) that inverts the state for each line through the terminal 36 to select the output of the adder 32 in the horizontal section T _HC on the input _side. Then, the output of the one line memory 34 is alternately switched to the input side.

Therefore, the original digital NTSC video signal can be obtained from the switch circuit 35 as shown in FIG.

The output of this switch circuit 35 is sent to the D-A converter 3.
7, the signal is returned to the original analog NTSC color video signal, and is led out to an output terminal 38.

As described above, in this invention, focusing on the fact that the sampling phase in sub-Nyquist sampling is a two-line period, the output of the sum and difference of data between two lines whose sampling phases are in phase is obtained, and this sum is and the difference output respectively
Since DPCM encoding is used, so to speak two-dimensional encoding is performed not only in the horizontal direction but also in the vertical direction, making it possible to improve the prediction characteristics of the DPCM encoder and enable significant compression. It is what it is. In other words, as mentioned above, DPCM
In predicting , not only horizontal correlation but also vertical correlation is used, so the prediction error becomes extremely small and the optimum prediction characteristics can be obtained. Therefore, compared to conventional DPCM encoding where only information within one line can be used for prediction, this invention uses information from the previous line with a strong correlation, resulting in significantly improved prediction. This enables data compression.

In addition, in conventional encoders, it is necessary to simultaneously predict both the luminance signal Y and chrominance signal C components using only sample values within one line, so the prediction function and nonlinear quantization characteristics of the predictor are restricted. However, according to the present invention, since the sum and difference between the two lines are divided into luminance signal Y, color signal C, etc., it is possible to use the optimum prediction function for each.

In other words, as mentioned above, it is possible to predict the brightness signal Y with good step response, and
As for the color signal C, it is possible to make a prediction that also takes into account the phase of the color subcarrier.

Therefore, there are very few large-level prediction error signals, and the precision of the nonlinear compressor can be increased accordingly.

In addition, according to this invention, since two-dimensional correlation can be effectively used, only one-dimensional prediction is possible, so overload noise, edge business, etc. DPCM
This greatly improves the conventional drawback that specific deterioration is easily noticeable on the screen.

In addition, in the case of this invention, since predictive encoding is performed only in units of two lines that are in phase, transmission is more efficient than conventional two-dimensional encoding, that is, a method that has a line memory in the predictive encoding loop. Another effect is that road errors are not propagated in the vertical direction.

Note that this invention is not limited to the case where the sampling frequency is 2f _SC and the sampling phase is inverted every two lines as described above;
Of course, it can also be applied to sub-Nyquist sampling in which the sampling phase is inverted every two lines in the case of 4f _SC and 3f _SC .

Furthermore, it goes without saying that the present invention is applicable not only to digital VTRs but also to all devices that transmit video signals in the form of digital signals, such as so-called digital television receivers.

[Brief explanation of drawings]

Figures 1 and 2 are diagrams for explaining sub-Nyquist sampling, Figure 3 is a system diagram of an example of a recording system when this invention is applied to a digital VTR, and Figure 4 is an example of a reproduction system. System diagram, Figure 5 is a diagram to explain the structure of the data, Figures 6 and 7 are diagrams showing the frequency spectrum of the sum and difference output between two lines, and Figure 8 is a diagram showing the nonlinear compression characteristics. The figures shown in FIG. 9 and FIG. 10 are diagrams for explaining the flow of data. 2 is an A-D converter, 4 is a sub-Nyquist sampling circuit, 13 and 14 are DPCM encoders, 1
32 and 142 are predictors.

Claims (1)

[Claims] 1. An NTSC signal in which the color subcarrier phase is inverted every line is subjected to sub-Nyquist sampling at a frequency such that the sampling phase is the same every two lines. A sum and a difference are determined, the sum output is sent to a DPCM encoder having a predictor of the first prediction function, and the difference output is sent to a DPCM encoder having a predictor of the first prediction function.
In a DPCM encoder with a predictor of prediction function,
DPCM encoding device adapted to be supplied respectively.