SU1239894A1 - Versions of method of coding and decoding television signal - Google Patents

Abstract

The invention relates to a television technique and provides a reduction in quantization errors. In the first embodiment, on the transmitting side, the array of initial samples is smoothed by two-dimensional low-pass filtering with bandwidths m and n times smaller than the nominal boundary frequencies, respectively, of the horizontal and vertical components of the television image. The array of samples is distributed over two-dimensional blocks of size m samples along the horizontal and n samples along the vertical. In each block, a sample count is allocated and the ei o value is transmitted (I I (L Vu RL-KE You o8 G | DU Vi o9 C. Fmg.

Description

sequence). Then, for each block, the difference between the original samples in the block and their smoothed values is determined and the signs of these differences are transmitted (sequence II). Next, averaged over the unit estimates of the amplitude of differences are determined and transmitted; their meaning (II sequence). Reverse operations are performed on the receiving side. The method is implemented. The invention relates to a technique of television, to methods of efficiently encoding a television signal for transmission over a communication channel or for recording, permitting to reduce the number of bits (bits) per image loss.

The purpose of the invention is to reduce encoding errors.

Fig. 1 shows the structural electrical circuit of the first variant of the device implementing the method of encoding and decoding the television signal; Fig. 2 shows the same second variant of the device implementing the method of encoding and decoding the television signal}, 3 g diagram, illustrating the method coding AND decoding of the television signal J in FIG. 4, the impulse response of the low-pass filter that implements the smoothing of the samples of the signal j in FIG. 5 is a diagram illustrating the interpolation of the samples for reception second side.

The method of encoding and decoding a television signal is carried out as follows.

In the first variant, on the transmitting side, the array of initial samples is smoothed by two-dimensional low-pass filtering with the horizontal and vertical bandwidths, respectively, w and n times smaller than the nominal boundary frequencies of the horizontal and vertical components of the initial sample. The array of samples is distributed over two-dimensional

vom, contain “on the transmitting side of the low-pass filter 1” sampler 2, delay block 3, subtraction unit 4, comparator 5, calculator 6, multiplexer 7, buffer drives 8 and 9, and on the receiving side - demultiplexer 10, LPF 11, calculator 12, block 13, which completes the operation EXCLUSIVE OR, delay blocks 14 - and 15, adder 16. 2 s, ffl, 5 nl.

blocks of size m counts horizontally and n counts vertically,

In each block, a reference sample is selected from the smoothed samples and its value is transmitted. For each block, the values of the signal of differences between the original samples in the block and their smoothed values are found. Generate and transmit the signs indicated

the differences over which the averaged (over the block) estimates of the amplitudes of the differences are found and transmit their values. At the receiving side, the transmitted reference samples are smoothed by

two-dimensional low-pass filtering with the same bandwidths as .and when smoothing the original samples on the transmitting side,

The second block of each block, using the signs of differences and averaged estimates of the amplitudes of differences, restores the values of the difference signals. Restore the array of signal samples, summing the smoothed sign in each block. Sample counts with reconstructed I difference signal values.

In the second embodiment, on the transmitting side, the selected reference samples

subjected to additional filtering with the same bandwidth as when filtering the original samples. After that, the value of the difference signal is found. In each block between the initial samples in the block and the samples obtained with additional filtering. All other operations on the transmitting and receiving sides are performed in the same sequence.

implications as for the first option,

As an example, consider the one-dimensional version with blocks of 4x1. In the diagram (Fig. 3q), the initial sequence of samples 1, the sequence of smoothed samples — P, obtained by filtering low frequencies with a passband of four times less than the nominal cut-off frequency of the horizontal components of the array of source samples, are completely absent in the smoothed sequence. -or signs of block construction. The dashed vertical lines here show the block boundaries, In addition, the blackened circles indicate the transmitted reference samples.

With additional filtering of reference samples, a sequence of samples is obtained, similar to that shown in the diagram (Fig. Za) in circles.

The diagram (Fig. 3) shows the signal values of the differences between the original samples and the smoothed values of the samples; in the diagram (Fig. 3 &) - the signs of the differences in the block Kaxj in the diagram (F1N, 32) - the average estimates of the amplitudes of the differences in the blocks in the daagram (Fig. 3a) are shown as on the receiving side from the reference samples using a filter , low frequencies with the same bandwidth as .and on the transmitting side, get smoothed samples,

The sample envelope thus obtained is similar to the bend obtained on the front side. The sampling smoothing procedure with the filtering power can be performed with any given degree of accuracy if the filter is closer to the ideal low-pass filter, since the conditions of the Kotelnikov theorem and the sample sequence will be more and more accurate. .

When restoring the signal samples by summing the smoothed samples with the average estimates of the amplitude differences taking into account the signs of these differences, the block structure does not occur.

2398AA 4

FIG. Figure 4 shows the impulse response of the low-pass filter, which implements horizontal smoothing for coding with blocks of size 4-1 by means of a convolution operation with a sequence of initial samples. In this case, the impulse response of the filter is a discrete function 10 with an envelope in the form of the principal positive lobe of the function

15

20

five

0

five

0

five

0

. -Mt, 7t - 4T 4T

where T is the step of the original samples.

If necessary, a better approximation to the ideal filter is required to increase the duration of the impulse response. In practice, the accuracy of the approximation is limited only by the permissible 1m object of calculations during filtering. However, analysis of the processes occurring during coding in the spectral region implies that the filtering with the one shown in FIG. 3 impulse response is sufficient to completely eliminate the block structure.

Fig. 5 shows how, at the receiving side, smoothing of reference samples can be performed using linear interpolation, which is implemented by known techniques and provides a first approximation to the required filtering. Here the counts No. 1 and 4 are reference ones, and No. 2 and 3 are interpolated. If necessary, increase the accuracy provided in this case (filtering linear interpolation can be replaced by parabolic or higher order interpolation. The same interpolation can be used to re-filter reference samples on the transmitting side according to the second embodiment of the method.

In the first embodiment (FIG. 1), the device on the transmitting side has a lower filter (LPF) 1., sampler 2, delay unit 3, subtracting unit 4, comparator 5, calculator 6, multiplexer 7, first buffer store 8, second buffer store 9, on the receiving side - demultiplexer 10, low-pass filter 11, calculator 12, block 13, first block

14 aderzhki, the second block 15 delay

adder 16..

The input of the device receives a sequence of initial signal samples. Then, the samples are filtered in the low-pass filter 1. In the sampler 2, rarer reference samples are selected from the smoothed sequence of samples. The initial sequence of samples through block 3 of the delay arrives at one input of the subtracting unit 4, the second input of which receives the smoothed sequence of samples. At the exit of the subtracting unit 4, the differences between the initial and smoothing and readings are formed. The delay unit 3 is used to compensate for the delay in the low-pass filter 1. With the help of a comparator 5, the difference signs are formed from the obtained differences. In the calculator 6, average estimates of the differences for each block are computed. As such estimates, the average or rms values of the amplitude of the differences in the blocks or other similar values can be used.

The reference samples, the signs of differences, and the averaged estimates of the differences arrive at the inputs of multiplexer 7 for the proportions of the signal in the transmitted signal. Before this, the reference samples are passed through the first buffer buffer 8, and the signs through the second buffer storage device 9.

When receiving, the time-compacted codes of the transmitted signals are sent to the demultiplexer 10, which distributes the signals in three directions. The reference samples arrive at the input of the low-pass filter 11. The averaged estimate of the differences arrive at the transmitter 12 - the transmitter of the values of the modules of the differences. In the simplest case, the value of the modulus of the difference can be equal to the transmitted average estimate of the differences; at the same time the calculator does not perform any operations. In block 13, which performs the EXCLUSIVE OR operation, the difference between the original and smoothed samples is restored from the modules of the differences and signs of the differences arriving at the second input of block 13 through the first block of delay 14, the restored differences pass through the second delay block 15, and adder 16 is summed with smoothed

j W 15 i 20 25

0 5

five

readings coming from the output of the LPI 1I. At the output of the device, reconstructed signal samples are formed.

According to the second variant (FIG. 2), the device comprises on the transmitting side of the low-pass filter 7j a sampler 18, an additional low-pass filter 19, the first buffer store 20, the delay block 21, the subtractor block 22, the comparator 23, the transmitter 24, the second buffer store 25, the multiplexer 26, on the receiving side of the demultiplexer 27, the low-pass filter 28, the transmitter 29, block 30. the first block 31 of the delay, the second block 32 of the delay, the adder 33,

The input of the device receives a sequence of initial samples of the signal. Then the samples are filtered. In the low-pass filter 17, in the sampler 18, rarer reference samples are selected from the smoothed sequence of samples. The initial sample sequence through the delay unit 21 is fed to one input of the subtracting: unit 22, to the second input of which the sequence of samples arrives after the additional radio, an additional low-pass filter 19, and the distortions introduced by the non-ideal low pass filter 28 at the receiving side are taken into account. At the output of the subtracting unit 22, differences are formed between the original and smoothed samples. The delay unit 21 compensates for the delay of samples in the low-pass filter 17. The comparator 23 of the obtained differences forms the signs of the differences. The calculator 24 averages the difference estimates for each block. Then the reference samples, difference signs, and averaged amplitude estimates are sent to multiplexer 26 to transmit the generated sample sequences. In this case, the reference samples pass through the first buffer accumulator 20, and the difference signs go through the second buffer accumulator.

drive 25. I

On the receiving side are sealed

over time, the codds of the transmitted signals fall on the demultiplexer 27, which distributes the signals in three directions. The reference samples arrive at the input of the low-pass filter 28, the averaged difference estimates go to the calculator 29, which calculates the value of the difference moduli. In block 30, the EXCLUSIVE operation is performed OR over modules of differences and signs of differences.

At this, the difference sign passes through the first delay unit 31, the differences between the original smoothed samples are restored and passed through the second delay unit 32, and in the adder 33 are added to the smoothed samples from the low-pass filter output 28, the recovered signal samples are formed at the device output,

formula of invention

I. A method of encoding and decoding a television signal, consisting in transmitting three sequences of samples, the first of which is formed by preprocessing the original samples,

including the distribution of samples in blocks with horizontal size m and vertical size n, and modifying onopHbix samples from preprocessed source samples, the second sequence of samples — by subtracting preprocessed samples from source samples and the subsequent formation of the polarity signal of the differences of the previously processed and source samples , the third sequence of samples is by forming the values of the difference signal of the previously processed values averaged over the blocks and receiving samples, and receiving three sequences of samples, followed by restoring the signal values of the differences of the previously processed and original samples and summing the recovered differences with the previously processed reference samples, characterized in that, in order to reduce coding errors during transmission, the preliminary processing of the original samples is carried out before distribution counting by blocks - by two-dimensional low-pass filtering with bands; transmissions are w and n times smaller than the nominal boundary frequencies, respectively, of horizontal and vertical connectors of the television image, and when received, the preprocessing is performed by two-dimensional filtering of low frequencies with bandwidths m m and n times smaller than the nominal boundary frequencies 39894 8

The values are respectively horizontal and vertical components of the television image. .

5 2. A method of coding and decoding a television signal, concluded in transmitting three sequences of samples, the first of which was formed by pre-processing of the original samples, including the distribution of samples in blocks with horizontal size m and vertical size n, and selection of reference samples of preliminary processed source counts, the second sequence is by generating a difference signal followed by generating a difference polarity signal, one third

20 sequence of samples - by forming the values of the difference signal averaged over blocks, and in receiving three sequences of samples with subsequent recovery

25 values of the difference signal and summation of the recovered values of the difference signal with the pre-processed reference samples, characterized in that

30 in order to reduce coding errors during transmission, pre-processing of the original samples is carried out before the samples are distributed in blocks by a two-dimensional filter, low frequency grating with bandwidths m and n times smaller than the nominal boundary frequencies, respectively, of the horizontal and vertical components of the television image, after the selection of reference samples, additional two-dimensional low-pass filtering is performed with bandwidths m and n times smaller than the nominal chad values of the frequencies co respectively horizontal and vertical components .televizi-. image, the signal of the difference is produced by subtracting an additional profile, tated reference samples, and when receiving, the preprocessing is carried out by two-dimensional filtering of low frequencies with passbands of r m and n times smaller than the nominal boundary frequencies, respectively

55

horizontal and vertical components of the television image.

 Lih

AND

I I t t I

Notebook g 3 ff - B

3

Jll -sx. L 0 L jr sr IT 2 g I I I g 2.

"Pmg. /

G 2 J t t And countdown 8 lacquer .S

Editor A.Sabo Zakae 34 and / 58

Compiled by T. Afanasyev

Tehred O.SOPKOKorrektor V. But ha

Circulation 624 Subscription

VNIIPI USSR State Committee

Inventions n discoveries 113035, Moscow, Zh-35, Raushsk nab. 4/5

Pro ductively

-shr ygr F meteorological advance, Uzhgorod. Design, .4

Claims (2)

Claim I. A method of encoding and decoding a television signal, which consists in transmitting three sequences of samples, the first of which is formed by pre-processing the original samples, • including the distribution of samples in blocks with horizontal size w and vertical size n, and extracting reference samples from pre-processed source samples , the second sequence of samples - by subtracting the pre-processed samples from the original samples and the subsequent generation of the field signal the differences of the pre-processed and initial samples, the third sequence of samples — by generating the block-averaged signal values of the differences of the pre-processed and the original samples, and in the reception of three sequences of samples with the subsequent restoration of the signal values of the differences of the pre-processed and original samples and summing the restored differences with the pre-processed reference samples, characterized in that, in order to reduce coding errors, when In the country, the preliminary processing of the initial samples is carried out before the distribution of the samples in blocks by two-dimensional low-pass filtering with bandwidths m and η times smaller than the nominal boundary frequencies of the horizontal and vertical components of the television image, and upon reception, the preliminary processing is carried out by two-dimensional low-pass filtering with bandwidths e w and a times smaller than the nominal boundary cha 94 8 Stot respectively horizontal and vertical components of the television image. 2. A method of encoding and decoding a television signal, which consists in transmitting three sequences of samples, the first of which is formed by pre-processing the initial samples, including the distribution of samples in blocks with horizontal size m and vertical size n, and extracting reference samples from pre-processed 15 initial · samples, the second sequence — by generating a difference signal followed by the formation of a difference polarity signal, the third 20 in series sampling rate - by generating the block signal values of the differences, and receiving three sequences of samples with the subsequent restoration of 25 values of the differences signal and summing the restored values of the differences signal with pre-processed reference samples, characterized in that, 30 in order to reduce encoding errors, pre-processing of the initial samples is carried out before the distribution of samples into blocks by two-dimensional low-pass filtering with bandwidths in w and η times smaller nominal cutoff frequencies respectively horizontal and vertical components of a television image ₄₀ zheniya after vydeleniyaopornyh samples performed more two-dimensional low-pass filtering with a bandwidth of m and η times smaller nominal boundary cha45 stot respectively horizontal and vertical components .televizi.onnogo Images; The difference signal is generated by subtracting additionally the profiles of ₅₀ reference samples, and upon reception, preliminary processing is carried out by two-dimensional low-pass filtering with pass bands of VT and η times less than the nominal ₅₅ boundary frequencies of the horizontal and vertical components of the television image, respectively. FIG. 2 3 ff reference 8 block 5