JPS6017184B2 - Predictive coding band compression device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a predictive coding band compression device, and more specifically, the prediction error between an input signal value and its predicted value is quantized, encoded and transmitted, and the transmission band width is compressed. This relates to improvements in equipment for

One method of compressing the transmission bandwidth when transmitting a coded signal is to predict the current signal value using a predetermined method based on past signal values, and then compare the predicted value with the actual signal value. There is a method of quantizing, encoding, and transmitting the difference, that is, the prediction error, and this is a predictive encoding band compression device.

FIG. 1 is a block diagram showing the construction of an example of a conventional predictive coding band compression device. 1 is a transmitting side device, 0' is a receiving side device, and the transmitting side device 1 is an analog/digital (A/D) device. ) converter 1, compressor 2, decompressor 3, predictor 4, texture calculator 5 for obtaining the difference between the output S of the A/D converter 1 and the output P of the predictor 4, that is, the prediction error x, and the predictor an adder 6 that obtains the sum R of the output P of 4 and the output Q of the expander 3, and the output Y of the compressor 2.
It consists of a transmitter 7 that transmits.

On the other hand, the receiving side device 0' includes a receiver 8, an expander 3', a predictor 4', an adder 6', a D/A converter 9, and a monitor 10.
It consists of As is clear from the figure, the system of expander 3, predictor 4, and adder 6 on the transmitting side is exactly the same as the system of expander 3', predictor 4', and adder 6' on the receiving side, and the transmission system If there is no noise at , the output R of adder 6 (transmission side) and the output R' of adder 6' (reception side) will be exactly the same. Now, as the past signal values that are the basis for predicting the current signal value mentioned above, it is also possible to obtain a high-order predicted value 4 by using values over the past several samplings of Doji-ized sampling. However, the configuration of the device becomes complicated and practical problems arise.

Therefore, in this case, a case will be explained in which only the sampling value of one step before is used. In the conventional device shown in FIG. 1, the predictor 4 is composed of a one-sample time delay circuit, and this device ultimately operates as a previous value predictive coding band compression device that uses the previous value as a predicted value.

Now, the output S obtained by A/D converting the input by A/D converter 1 changes to a range of ~63 (6 bits), and the characteristics of compressor 2 and expander 3 are as shown in Figure 2 and the table below. Suppose that

That is, the value of the difference x n between the signal Sn and the predicted value Pn (=previous value Rn-,) is 0≦×n<2, 2≦×n5, 5≦X
Compressor 2 depending on n<10 or 10xnS63
The output Yn of is L1 (000), L2 (001), L3 (
010) or L4 (011), and the output Qn of the expander 3 is 1.0 (000010) and 3.0, respectively.
((000110), 70 (001110), or 13.0 (011010).

However, if the value of difference x n is negative, the first bit of the binary code of outputs Yn and Qn becomes "1". Note that in this), the subscript n represents the time of sampling, and indicates the value at the nth sampling time. Now, an example of an operating waveform will be described with reference to FIG. 3 when the waveform of the output S of the A/D converter 1 is a waveform with an amplitude of 20 that increases or decreases in step size 4 as shown as S in FIG. 3.

Such a waveform has a low frequency compared to the sampling frequency, and the amplitude is sufficiently small compared to the maximum amplitude 127, and is considered to occur quite frequently in general image transmission. First, at the sample time to,
If the output signal Ro of the adder 6 is 0, then at the next sample time t, the output signal P of the predictor 4 becomes 0 (P
n=Rn-,), at that point the signal S,=4
Therefore, the output signal of the subtracter 5 becomes X,=S,-P,=4.

From the companding frame properties shown in FIG. 2 and the table above, the companding output Q,=3, and since the signal Rn:Pn+Qn, R,=3.
At the next sample time, P2=R,=3 and S2=
8, so X2 x S2 - P2 = 5, and from the companding barrel property, Q2 = 7, and therefore R2 = P2 + Q2 = 10.
Thereafter, signals R, P, and Q shown in FIG. 3 are obtained in the same manner. The output signal Y of the compressor 2 of such a system is sent to the receiving side 0 through the transmitter 7. This signal Y is only 3 bits of information as shown in the table above, and on the receiving side B, after receiving it with the receiver 8, a signal Y' corresponding to this signal Y is sent.
By supplying the signal R' to the same system of expander 3', predictor 4', and adder 6' as in the transmitter 1, a signal R' exactly the same as the signal R in the transmitter 1 can be obtained at the output of the adder 6'. . The difference between this signal R' and signal S is shown in Figure 3 as "R-S".
Shown as In this way, the transmission band width is compressed.
By the way, if you try to increase the compression effect by reducing the number of bells as much as possible by quantizing the quantization described above, an error (noise component) will occur due to the quantization, and as a result, the signal-to-noise ratio (S/N ratio) of the signal will deteriorate. do.

Furthermore, when the transmitted data is an image signal, there are various causes of deterioration specific to images in addition to deterioration of the average S/N ratio. Part 1
One of them is deterioration due to ``gradient overload,'' which appears as ``blur'' on the contours of the reproduced image. This occurs because the response of the band compression device to sudden large-amplitude changes in the input is insufficient, and this occurs because the maximum value of the quantization bell is small relative to the maximum amplitude of the prediction error. FIG. 4 is a waveform diagram for explaining the operation of the conventional device in response to such a sudden large-amplitude change in input. A case in which the signal changes in a pulse-like manner to the full 63 will be explained.

First, at the sampling time to, the output signal R of the adder 6
If o is 0, the output signal of the predictor 4 is P,:0 at the next sample time, and at that point, the signal S,
=63, so the output signal of the subtractor 5 is X, =S, -P,
=63. From the expansion characteristics shown in FIG. 2 and the table above, the companding output Q,=13, and the signal R,=P,Q,=13. At the next sample time, P2=R,=13,
Since S2=63,
It becomes 6. Furthermore, at the time of the next sample, P3=R
2=20 S3=63, so X3=S3-P3:3
7, Q3=13, so R3=P3+Q3=39. Similarly, the signals R and P shown in FIG. 4 are obtained, but as is clear from the figure, a large error occurs when the input signal changes. As explained above, the conventional device using the simple front pupil prediction method shown in FIG. 1 generates errors (noise components) as shown in FIG.
An input signal having a sudden large amplitude change as described with reference to the figure will cause a large error, which corresponds to such a case in the contour portion of an image signal, causing contour blur in the reproduced image.

To alleviate such problems, there are methods that change the prediction formula according to the instantaneous value of the input, or variable quantization methods that keep the prediction formula constant and quantize it depending on the magnitude of the prediction error to change the bell value. have been proposed, but both methods require complicated equipment and the need to transmit predictive formulas or quantized information with varying bell values to the receiving side, so the effect of Obijo compression cannot be maintained. It had the disadvantage of being exposed.

This invention was made in view of the above points, and includes:
By using both the signal reproduction value before sampling and the prediction error at that time, prediction accuracy is improved and the average S/N
It is an object of the present invention to provide a predictive coding band compression device that increases the ratio, speeds up the response to sudden large-amplitude changes in the input, and reduces the influence of "gradient overload."

FIG. 5 is a block diagram showing the configuration of an embodiment of the present invention, except that the configuration of the predictors 4a, 4a' is different from that of the predictors 4, 4' of the conventional device in FIG. 1. This is the same as the conventional device. The predictor 4a of this embodiment contains the output Qa from the expander 3 in addition to the output Ra from the adder 6. FIG. 6 is a block diagram showing an example of the configuration of the predictor 4a used in the present invention, in which 11 is a one-sample time delay circuit similar to that used as the predictor of the slave device, 12 is a coefficient circuit, and 13
is an adder. The coefficient circuit 12 multiplies the companding output Qa by a positive predetermined coefficient Q to obtain QQa, and is generally configured by a combination of adders and subtracters, but when the coefficient Q is 2p (p
is an integer), it is simple in a binary system, and each bit needs to be shifted by P bits. Figure 7 shows Q=2-1
FIG. 2 is a configuration diagram of the coefficient circuit 12 in the case of FIG. In short, in the predictor 4a of the present invention, the Qa value (prediction error x a) of the previous sample is added to the Ra value of one sample before the compressor 2 and the expander 3. It is the value obtained by adding the value pressed by ) multiplied by the coefficient .

This serves to improve the prediction accuracy of the predictor 4a. For example, when the input increases (or decreases) over several samples, the prediction error X is always positive (or negative) in the conventional simple previous value prediction. In such a case, the prediction accuracy can be improved by selecting an appropriate value using the predictor 4a of the present invention and making an appropriate correction to the Ra value one sample time ago. In this way, if the prediction accuracy is improved,
The prediction error X becomes smaller, and the value Q obtained by companding it also takes a smaller value. Generally, as shown in the example shown in Figure 2, the companding characteristic is set densely in the range where the difference signal input, that is, the prediction error, is small, and coarsely in the large range. The smaller the prediction error X is, the less it becomes. That is, as the prediction accuracy improves, the average S/N ratio of the reproduced image increases. On the other hand, the response to a step-like change with a large amplitude is quantized to the maximum and the bell value is (10Q) times that of the conventional one, so that the influence of "gradient overload" can be reduced. FIG. 8 is a waveform diagram showing the operation of this embodiment device with Q=0.5 for an input waveform similar to that of FIG. 3.

At the time of sampling to', the output signal Ra of the adder 6 is 0.5 and the companding output Qao is 1, so from the predicted value Pa,=RSQQa, at the time of sampling, P mountain=0. becomes. Since the signal S,=4 at this point, the output signal of the subtracter 5 becomes Xa,=S,-Pa,=4, and from the companding characteristics, Q. ,=． It becomes 3. And signal R illusion = Pa,
Since it is a 10Q mountain, Ra,=3. Next sample time t
2 has Pa2=Ra, 1QQa, :4.5, and S2
=8, so Xa2=S2-P42=3.5, and from the companding characteristics, Q=3. Therefore, Ra2=Pa2+Qa2=7.5. Thereafter, signals Ra, Pa, and Qa shown in FIG. 8 are obtained in the same manner. Then, the reception output Ra' at the receiver side becomes the same as the above signal Ra. The difference between the received output Ra′ and the signal S, that is, the noise component, is “
The improvement in the S/N ratio can be understood as shown by "Ra-S" compared to FIG. 3 for the conventional device. FIG. 9 is a waveform diagram showing the operation of the same device as in FIG. 4 in response to a sudden large-amplitude change input for the same embodiment.

Assuming that the signals Rao = 0 and Qao = 0 at the sampling time to, the predicted value P mountain at the sampling time t =
0, and at this time S,:63, so the signal xa,=
63 Companding input Qa, = 13 Therefore Ra, = Pa, 10Qa
,=13. Tsuzu〈At sample time t2, Pa2il
9.5, since it is still S2i63, X2=43.5, Q
a2=13, and Ra2=32.5. Thereafter, signals Ra, Pa, and Qa shown in FIG. 10 are obtained in the same manner. The signal output R. obtained in this way. (Reception output R
Comparing ``a'' (corresponding to this) with that of the conventional device shown in Fig. 4, the size of one step has increased approximately 1.5 times, and the response to input is that much faster, reducing the problem of ``gradient overload.'' It can be seen that the impact is reduced. Note that in the above embodiment, the coefficient value Q of the coefficient circuit 12 in the predictor 4a
Although the case where Q is 0.5 has been described, in general, an appropriate value of the coefficient Q may be selected depending on the characteristics of the signal source.

As described in detail above, in this invention, the prediction error Xn between the sampled value Sn obtained by sampling the input signal and its predicted value Pn is compressed and sent to the receiving side, and the compressed signal Yn is The predicted value Pn+ at the time of the next sample is determined based on the signal reproduction value Rn obtained by expanding the compressed/expanded signal Qn and adding the predicted value Pn to it.
, the predicted value P at the time of the next sample is
Since n+, is made to be the sum of the signal reproduction value Rn and the value obtained by multiplying the compressed and expanded signal Qn by a positive predetermined coefficient Q, the conventional signal reproduction value Rn can be simply converted into the next predicted value Pn+, This is an excellent prediction code that improves prediction accuracy, increases the average S/N ratio, and reduces the effects of "gradient overload" with just the addition of a small number of component circuits, compared to the prediction method described above. A compact compression device is obtained.

[BRIEF DESCRIPTION OF THE DRAWINGS] Fig. 1 is a block diagram showing the structure of an example of a conventional predictive coding band compression device, Fig. 2 is a characteristic diagram showing the characteristics of a compression/expansion device, and Figs. The figure is an operation waveform diagram for explaining the operation of this conventional device, FIG. 5 is a block diagram showing the configuration of an embodiment of the present invention, and FIG. 6 is a block diagram showing an example of the configuration of the predictor used in the present invention. Figure 7 is a block diagram showing an example of a coefficient circuit used in this predictor, Figures 8 and 9 are diagrams showing an example of a coefficient circuit used in this predictor.
The figure is an operational waveform diagram for explaining the operation of this embodiment. In the figure, 1 is an A/○ converter, 2 is a compressor, 3 is an expander, 4 and 4a are predictors, 5 is a subtracter, 6 is an adder, 7 is a transmitter, 8 is a receiver, and 9 is a A D/A converter, 11 a one-sample delay circuit, 12 a coefficient circuit, and 13 an adder. Note that the same reference numerals in the figures indicate the same or corresponding parts. Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9

Claims (1)

[Claims] 1. Sampling value S obtained by sampling the input signal
The prediction error X_n between __n and its predicted value P_n is compressed and sent to the receiving side, and the compressed signal Y_
n is expanded to obtain a compressed/expanded signal Q_n, and the predicted value P_n_+_1 for the next sample is obtained based on the signal reproduction value R_n obtained by adding the predicted value P_n to the compressed/expanded signal Q_n. Predicted value P_n_+ at sample time
_1 is the signal reproduction value R_n and the compressed/expanded signal Q_n.
1. A predictive coding band compression device characterized in that the value is the sum of a value obtained by multiplying .alpha. by a predetermined positive coefficient α. 2. The predictive coding band compression device according to claim 1, wherein the value of the coefficient α is 0.5.