JPS6262111B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a signal preprocessing method for improving prediction accuracy in a predictive coding method for image signals such as television signals.

A predictive coding method is known as an efficient coding method for television signals. This encodes the difference between the input signal value and its predicted value, that is, the prediction error. In order to increase the efficiency of this predictive coding, it is necessary to reduce the prediction error, that is, to improve the prediction accuracy.
One method for this purpose is to perform preprocessing on the input signal to modify the signal before predictive encoding, thereby improving prediction accuracy. This is called a preprocessing method. Conventionally, as this preprocessing method, a method is known in which signals are smoothed between lines, fields, or frames of a television signal to strengthen the correlation between pixels. However, this conventional method cannot be applied as a high-quality encoding method because image quality deteriorates when the outline of the image in the diagonal direction or fine patterns move.

In order to solve the above-mentioned problems found in the conventional method, the present invention adaptively performs preprocessing in accordance with the local characteristics of image signals. DESCRIPTION OF THE PREFERRED EMBODIMENTS The predictive coding preprocessing method of the present invention will be explained in detail below based on an embodiment thereof using the drawings.

FIG. 1 shows the configuration of a first embodiment of the present invention, in which a preprocessing circuit A and a predictive encoding circuit B are connected in series. Preprocessing circuit A
The input terminal t ₀ of is connected to one of the input terminals of the subtracter 11 and the adder 14, respectively, and is further connected to the delay circuit 1.
It is connected to the second input terminal. This delay circuit 1
The output terminal of the subtracter 11 is connected to the other input terminal of the subtracter 11, and the multiplier 13 is connected between the output terminal of the subtracter 11 and the other input terminal of the adder 14. The output terminal of this adder 14 is the input terminal _t3 of the predictive encoding circuit B connected to the next stage. An input signal X(t) is applied to the input terminal _t0 of the preprocessing circuit A having such a configuration, and this is delayed by a period T in the delay circuit 12 to become a signal X(t-T). -T) and X(t) are calculated by the subtracter 11. Further, the output signal of the subtracter 11 is multiplied by α in the multiplier 13 and is applied to one of the input terminals of the adder 14.

The input terminal t ₃ of the predictive encoding circuit B is connected to the input terminal of a subtracter 15 , and the output terminal of this subtracter 15 is connected to the input terminal of a quantization encoder 16 . An adder 17 is connected between the prediction error terminal t ₁ of this quantization encoder 16 and the other input terminal of the subtracter 15.
A delay circuit 18 is connected via. Further, the output terminal of this delay circuit 18 is connected to the other input terminal of the adder 17. The encoded output will be available at the other output terminal t ₂ of the quantization encoder 16.

Here, the input signal X(t) to the preprocessing circuit A, the output signal of the delay circuit 12 and the signal X(t-T) which is a signal before the input signal X(t) by a time T, and the multiplier 1
The input signal X ₁ (t) of No. 3 and the output signal X ₂ (t) of the multiplier 13 are signals from various parts of the preprocessing circuit A, but both represent sample values of the signals. Similarly, input signal to predictive encoding circuit B
X'(t), the output signal of the delay circuit 18 and the input signal
The encoded/decoded signal X'(t-T), which is a signal that is a time T before X'(t), is a signal from each part of the predictive encoding circuit B, but both are sample values of the signal. shall represent.

Next, the present invention will be described in the case of inter-field predictive coding of an NTSC color television signal.
In this case, a 1-field delay memory, more precisely a 262-line delay memory, is used as the delay circuits 12 and 18 in FIG.

First, the subtracter 11 inputs the input signal X(t) to the preprocessing circuit A and the input signal
Subtraction is performed with the previous signal X(t-T) by a field time) to obtain X ₁ (t)=X(t-T)-X(t). In the multiplier 13, this signal X ₁ (t)
is multiplied by a coefficient α, and X ₂ (t)=α{X(t-T)-X(t)} is obtained as a signal at its output end. Here, the value of α is given for each sample by the following equation according to the absolute value |X(t-T)-X(t)| of the input signal X ₁ (t) to the multiplier 13.

However, in equation (1), T _H is a predetermined threshold value. Therefore, the output signal X ₂ of the multiplier 13
(t) is given by the following equation.

Next, in the adder 14, this signal X ₂ (t) is added to the input signal X(t) to the preprocessing circuit A, and the output signal X'(t) of the preprocessing circuit is given by the following equation. .

This signal X'(t) becomes an input signal to the next inter-field predictive coding circuit B. The subtracter 15 subtracts the input signal X'(t) from the predicted value X'(t-T), which is the output signal of the delay circuit 18. encoded. The adder 17 adds the quantized prediction error obtained from the prediction error terminal t ₁ of the quantization encoder 16 and the predicted value, and the result becomes an input signal to the delay circuit 18 .

FIG. 2 shows the arrangement of lines (scanning lines) on a television screen and the color subcarrier phase of each line. The n-th line and the n-262nd line 262 lines before it are spatially adjacent to each other on the screen, and the color subcarrier phases of both lines match. Further, sample point B corresponding to signal value X(t-T) is located directly below sample point A corresponding to signal value X(t). Therefore, in inter-field predictive coding of an NTSC color television signal, the value at point A is predicted by using the value at point B, which has already been encoded and decoded.

Considering equation (3) in Figure 2, if the difference between the signal amplitude values at point A and point B is less than a certain value, the signal value at point A is replaced with the average value of the two, and the above difference becomes a constant value. If it exceeds , it means that the signal value at point A is preserved as is. Because replacing the signal value X(t) at point A with the average value of itself and the signal value X(t-T) at point B will strengthen the correlation between pixels between points A and B. , such signal preprocessing can improve the accuracy of inter-field prediction. On the other hand, such averaging processing of signals between fields causes a decrease in vertical resolution and blurring of moving parts.

Incidentally, such deterioration in image quality due to a decrease in vertical resolution or blurring of moving parts is only detected by human vision when the difference in signal amplitude between points A and B is larger than a certain level. According to experiments, for a 1Vpp decoded image signal (including a synchronization signal), the difference in amplitude between points A and B is 8/256V ~
If the voltage is below about 12/256V, no deterioration in image quality will be detected. Therefore, as in this method,
The signal is averaged between fields only when the difference between (t) and X(t-T) is less than a certain value, and no averaging is performed when the difference exceeds a certain value. This makes it possible to improve prediction accuracy without causing deterioration in image quality. Of course, the degree of improvement in prediction accuracy is somewhat smaller than in the conventional method, which constantly performs averaging processing, but experiments have shown that the difference is small.

FIG. 3 shows the configuration of a second embodiment of the present invention, in which the same parts as those in FIG. 2 are designated by the same symbols. In this second embodiment, the delay circuit 12 of the preprocessing circuit A is omitted in the circuit of the first embodiment, and the delay circuit 18 of the predictive encoding circuit B is
The output terminal of the preprocessing circuit A-1 is connected to one of the input terminals of the subtracter 11 of the preprocessing circuit A-1.

Here, the input signal X to the preprocessing circuit A-1
(t), input and output signal of multiplier 13 X ₁ (t)
and X ₂ (t), input signal of predictive encoding circuit B
X'(t), the output signal X'(t-T) of the delay circuit 18 which delays this input signal X'(t) by time T.
Let each represent a sample value of the signal.

Next, as in the first embodiment, the present method will be explained for the case of inter-field predictive coding of an NTSC color television signal. In this case, the use of a 1-field delay memory, more precisely a 262-line delay memory, as the delay circuit 18 is the same as in the first embodiment. First, the subtracter 11 inputs the input signal X(t) to the preprocessing circuit A-1 and the output signal of the delay circuit 18.
Subtraction is performed between X'(t-T) and _X1 '(t)=X'(t-T)-X(t). In the multiplier 13, this signal X ₁ '(t)
is multiplied by a coefficient α, and X ₂ '(t)=α{X'(t-T)-X(t)} is obtained as a signal at its output end. Here, the value of α is given for each sample according to the absolute value |X'(t-T)-X(t)| of the input signal X ₁ '(t) to the verticalizer 13 as follows.

However, in equation (4), T _H is a predetermined threshold value. Therefore, the output signal of multiplier 13
X ₂ '(t) is given by the following equation.

Next, in the adder 14, this X ₂ '(t) and the input
(t), and the output signal X'(t) of the preprocessing circuit A-1 is given by the following equation.

This signal X'(t) becomes an input signal to the inter-field predictive coding circuit B, and the same coding process as in the first embodiment is performed. Equation (6) is obtained by replacing X(t-T) with X'(t-T) in Equation (3), and X'(t-T) is replaced by X(t-T).
quantization noise is added to the quantization noise. As can be seen from equation (6), also in this second embodiment, the signal value
(t) and the signal value Since this is not performed, it is possible to improve prediction accuracy without causing deterioration of image quality.

FIG. 4 shows the configuration of a third embodiment of the present invention, which is similar to the first embodiment shown in FIG. The configuration is exactly the same as that of the embodiment. At this time, the output signal X ₁ (t) of the subtracter 11 is X ₁ (t) = X (t - T) - X (t), and the output signal X ₂ (t) of the adder 13 is X ₂ (t) = α {X _H (t-T)-X _H (t)}. However, X _H (t) and X _H (t-T) represent high frequency components of the signal X(t) and the signal X(t-T), respectively. Output signal of preprocessing circuit A-2
X′(t) becomes X′(t)=X(t)+α{X _H (t−T) −X _H (t)}, where the value of coefficient α is the absolute The value |X _H (t−T)−X _H (t)| is given for each sample as follows.

Therefore, the output signal of preprocessing circuit A-2
Equation (8) is given as X'(t).

As is clear from equation (8), X _H (t) and X _H (t-
If the difference in T) is below a certain value, the low frequency component is
Using the low frequency component (X(t)-X _H (t)) of (t) as is, the inter-field average is calculated between _{X H (} t) and The resultant data is subjected to encoding processing and is input to a predictive encoding circuit. On the other hand, if the difference between X _H (t) and X _H (t-T) exceeds a certain value, X(t) is input as is to the predictive encoding circuit.

In this embodiment as well, the input signal is corrected only when the difference between X _H (t) and X _H (t-T) is less than a certain value, so the accuracy of inter-field prediction is maintained without deteriorating image quality. can be improved. Furthermore, in this embodiment, signal averaging processing is performed only on high-frequency components, and since human visual characteristics deteriorate the higher the frequency components are, the impact of such image modification on image quality is limited. This is even less than in Examples 1 and 2.

FIG. 5 shows the configuration of a fourth embodiment of the present invention. The configuration is exactly the same as that of the second embodiment shown in the figure. Preprocessing circuit output in this case
X'(t) is given by equation (9) in the same way as in the previous embodiment.

X'(t)=X(t)+α{X _H '(t-T) -X _H (t)} (9) Here, X _H '(t-T) is X'(t −
T) represents the high frequency component. Also, the coefficient α is |X
It is given by the following formula for each sample depending on the value of _H ′(t-T)−X _H (t)| Therefore, the output signal X'(t) of preprocessing circuit A-3 is
It is given by equation (11).

Equation (11) converts X _H (t-T) to X _H '(t-T) in Equation (8).
, and X _H '(t-T) is X _H
(t-T) plus quantization noise.
As is clear from equation (11), X _H (t) and X _H '(t-
If the difference in T) is below a certain value, the low frequency component is
Using the low frequency component (X(t) - X _H (t)) of (t) as is, only the high frequency component is averaged between fields between X _H (t) and X _H '(t - T). It is processed and input to the predictive encoding circuit. On the other hand, if the difference between X _H (t) and X _H '(t-T) exceeds a certain value, X(t) is input as is to the predictive encoding circuit. In this embodiment as well, signal correction is performed only when the difference between X _H (t) and X _H ′ (t-T) is less than a certain value, so it is possible to improve prediction accuracy without deteriorating image quality. can.

FIG. 6 shows an example of experimental results in the second embodiment of the present invention, with I ₂ when preprocessing is performed on each image and I ₁ when no preprocessing is performed.
The prediction error enloby for each is shown.
By performing the preprocessing of this method as shown in the figure, the enloby can be reduced by about 30%. Furthermore, when the image corrected by this preprocessing was compared with the original image, no deterioration in image quality was detected at all.

Although inter-field predictive coding has been described in the embodiments described above, the present invention is not limited thereto, and can also be applied to inter-frame predictive coding and inter-line predictive coding. In the case of interframe predictive coding, the delay time T of the delay circuits 12 and 18 is set to one frame time, and in the case of interline predictive coding, T is set to one line time. In addition, in the explanation of the above embodiment, the value of α is 1/2.
Although the method of adaptively selecting two values, 0 and 0, has been described, the present invention is not limited to this, and it is also possible to select three or more values as the value of α, and that value can also be implemented. 0 and 1/ as in the example case
Of course, the value is not limited to 2, and other values are also possible.

As explained above, the present invention performs signal preprocessing in response to the local characteristics of the image signal, without causing deterioration in image quality such as a decrease in resolution or blurring of moving parts. Since prediction accuracy can be improved, there is a great advantage in improving coding efficiency in high-quality coding.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing the configuration of a first embodiment of the present invention, Fig. 2 is a diagram showing the line arrangement and color subcarrier phase on a television screen, and Fig. 3 is a diagram showing the configuration of a second embodiment of the invention. FIG. 4 is a diagram showing the configuration of a third embodiment of the present invention, FIG. 5 is a diagram showing the configuration of the fourth embodiment of the invention, and FIG. 6 is a diagram showing the characteristics of the present invention. It is a figure showing an example. 11: Subtractor, 12: Delay circuit, 13: Multiplier, 14: Adder, 15: Subtractor, 16: Quantization encoder, 17: Adder, 18: Delay circuit, 1
9: High pass filter.

Claims (1)

[Claims] 1. In a predictive coding method for image signals,
When (t) is a signal value at time t, X(t-T) is a signal value past time t by an integral multiple of the repetition period of the image signal, and α is a number satisfying 0≦α≦1, X in advance before predictive encoding processing
(t), X'(t)=X(t)+α{X(t
-T)-X(t)}
Let X'(t) be the input signal value to the predictive encoding circuit, and let the value of α in the above calculation be X(t) and
Pre-processing for predictive encoding characterized by changing the absolute value of the difference from (t-T) to a smaller value when it is larger than a predetermined value, and to a larger value when it is smaller than the predetermined value. method. 2 Signal value X(t) at time t and signal value X past time T by an integral multiple of the repetition period from time t
(t-T) for these high-frequency components and change α according to the absolute value of the difference. .