JPH03273762A - Bandwidth compression method - Google Patents

Abstract

PURPOSE:To always execute a processing using a predictive function suitable for each picture and picture element and to improve compression efficiency by calculating the estimated value of the current picture element from the predictive function of a coefficient with the absolute value of estimated error to the preceding picture element, which is already encoded, as a minimum, CONSTITUTION:A calculation processing part 15 for the optimum predictive function calculates the predictive function with the absolute value of the estimated error as the minimum in respect to the restorage signal of the preceding picture element, which is already encoded, to be outputted from an addition processing part 14. A predictive processing part 16 forms an estimation signal from the predictive function calculated by the calculation processing part 15 for the optimum predictive function in the preceding step and supplies the formed signal to a difference processing part 11 and the addition processing part 14. Then, the estimated value of the current picture element is calculated from the predictive function of the coefficient with the absolute value of the estimated error to the preceding picture element, which is already encoded, as the minimum. Thus, the processing is enabled while using the predictive function suitable for each picture and picture element at all times and the compression efficiency can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a band compression method for band compression by predictively encoding an image signal.

[Prior art] When band-compressing a wider-band image signal than conventionally, the fifth method is used.
Predictive encoding processing as shown in the figure was performed.

In FIG. 5, 1 is a difference processing section, 2 is a nonlinear quantization processing section, 3 is an encoding processing section, 4 is an addition processing section, and 5 is a prediction section.

The processing operation shown in FIG. 5 will be explained below.

In FIG. 5, a sampled image signal is manually input to the difference processing unit 1, and the error IA burying unit 1 combines the input sampled image signal with a prediction signal output from the prediction unit 5, which will be described later. The difference between both signals is calculated, and the difference value between both signals is supplied to the nonlinear quantization processing section 2 as a prediction error signal.

The nonlinear quantization processing section 2 nonlinearly quantizes the manually input prediction error signal, and supplies the signal to the encoding processing section 3 and addition processing section 4 .

Then, the encoding processing unit 3 converts the nonlinear quantized prediction error signal into a code word suitable for electronic transmission or recording, and outputs it onto a transmission path.

The nonlinear quantized prediction error signal output from the nonlinear quantization processing section 2 is also supplied to the addition processing section 4, and the addition processing section 4 combines the nonlinear quantization prediction error signal with the nonlinear quantization prediction error signal, which will be described later. A restored signal is formed by adding the prediction signal outputted from the prediction unit 5 and is supplied to the prediction unit 5.

The prediction unit 5 has a prediction function that forms a prediction signal for the image signal to be encoded next based on the restored signal output from the addition processing unit 4, and calculates the difference between the formed prediction signal as described above. Processing unit 1. It is supplied to the addition processing section 4.

As described above, band compression by predictive coding processing as shown in Figure 5 concentrates the appearance frequency distribution of prediction errors near zero prediction error, as shown in Figure 6, and reduces the absolute value of the prediction error. If the absolute value of the prediction error is small, fine quantization is performed, and if the absolute value of the prediction error is large, coarse quantization is performed, thereby compressing the band with less information loss. The compression effect improves as the prediction error is concentrated in the vicinity, so in order to concentrate the appearance frequency distribution of prediction errors in the vicinity where the prediction error is zero, the prediction function in the prediction unit 5 is adapted to the correlation characteristics of the image. Must be set.

[Problems to be Solved by the Invention] However, in the predictive coding process as shown in FIG. Since the average value is set, it is not optimal for each individual image, and it is also not optimal for the local correlation characteristics of the image, so information loss does not necessarily occur. There have been problems in that it is not possible to perform band compression with a small amount of noise, resulting in deterioration of the image signal, and it is not possible to improve the compression efficiency of the image signal.

In order to solve the above-mentioned problems, the present invention focuses on the correlation that partial image regions have, for example, the high correlation between adjacent pixels, and By finding the coefficients of the optimal prediction function and processing the current pixel with the prediction function of these coefficients, it is possible to always process using the prediction function that is suitable for each image and pixel.
The purpose of this invention is to provide a band compression method that can improve compression efficiency.

[Means for solving the problem] The band compression method of the present invention is a method of performing band compression by predictively encoding an image signal, and is a method that performs band compression by predictively encoding an image signal. This method is characterized in that the predicted value of the current pixel is obtained by a predictive function of coefficients.

[Operation] By the above method, processing can always be performed using a prediction function that is suitable for each image and pixel, and compression efficiency can be improved.

[Example code] Hereinafter, the present invention will be explained using examples of the present invention.

FIG. 1 is a diagram showing the principle of predictive encoding processing to which the present invention is applied, as an embodiment of the present invention.

In FIG. 1, 11 is a difference processing section, 12 is a nonlinear quantization processing section, 13 is an encoding processing section, 14 is an addition processing section, 15
1 is an optimal prediction function calculation processing unit, and 16 is a prediction processing unit.

The processing operation shown in FIG. 1 will be explained below.

In FIG. 1, a difference processing unit 11 calculates the difference between a sampled image signal and a prediction signal output from a prediction processing unit 16 (described later), and the difference value between both signals is nonlinearly quantized as a prediction error signal. The signal is supplied to the processing section 12.

The nonlinear quantization processing section 12 nonlinearly quantizes the input prediction error signal and supplies the signal to the encoding processing section 13 and the addition processing section 14 .

Then, the encoding processing unit 13 converts the nonlinear quantized prediction error signal into a code word suitable for electronic transmission or recording, and outputs the code word onto a transmission path.

On the other hand, the nonlinear quantized prediction error signal output from the nonlinear quantization processing section 12 is also supplied to the addition processing section 4, and the addition processing section 14 converts the nonlinear quantization prediction error signal into a nonlinear quantized prediction error signal, which will be described later. A restored signal is formed by adding the predicted signal output from the prediction processing section 16, and is supplied to the optimum prediction function calculation processing section 15.

The optimal prediction function calculation processing unit 15 calculates the restored signal of the already encoded previous pixel output from the addition processing unit 14.
The prediction function that minimizes the absolute value of the prediction error is calculated, and the prediction processing unit 16 further forms a prediction signal from the prediction function obtained by the optimal prediction function calculation processing unit 15 in the previous stage, and calculates the difference as described above. Processing unit 11. It is supplied to the addition processing section 14.

FIG. 2 is a diagram showing the principle of decoding processing (band expansion processing) corresponding to the predictive encoding processing shown in FIG. 1.

In FIG. 2, 21 is a decoding processing section, 22 is an addition processing section, 23 is an optimal prediction function calculation processing section, and 24 is a prediction processing section.

The processing operation shown in FIG. 2 will be explained below.

In FIG. 2, a coded signal supplied from a transmission path is input to a decoding processing unit 21, the coded signal is converted into a prediction error signal, and an addition processing unit 22 converts the coded signal into a prediction error signal.
is supplied to

The addition processing section 22 restores the sampled image signal by adding the prediction error signal supplied from the decoding processing section 21 and the prediction signal output from the prediction processing section 24, which will be described later, and outputs it.

On the other hand, the restored sampled image signal output from the addition processing section 22 is also supplied to the optimal prediction function calculation processing section 23, and the optimal prediction function calculation processing section 23 outputs the restored sampled image signal from the addition processing section 22. The prediction function that minimizes the absolute value of the prediction error is calculated for the restoration signal of the previous pixel that has already been restored. A prediction signal is formed from the function and is supplied to the addition processing section 22 as described above.

Next, detailed processing operations of the optimum prediction function calculation processing section 15.23 and the prediction processing section 16.24 in the predictive encoding processing shown in FIG. 1 and the decoding processing shown in FIG. 2 described above will be explained.

FIG. 3 is a diagram showing an embodiment of the optimal prediction function calculation processing section and prediction processing section in FIGS. 1 and 2.

In FIG. 3, 31 is a first predicted signal forming section, 32 is a second predicted signal forming section, 33 is a third predicted signal forming section, and 34
- 36 are predicted signal holding units, 37 to 39 are difference processing units, 4
0 is a discrimination section, and 41 is a prediction processing section.

The processing operation shown in FIG. 3 will be explained below.

In FIG. 3, the first predicted signal forming section 31, the second predicted signal forming section 32, and the third predicted signal forming section 33 each have coefficients of different prediction functions, and the addition processing section 14.22
The prediction signals formed in each prediction signal forming section are supplied to the prediction signal holding sections 34 to 36 and the prediction processing section 41, respectively.

Note that the coefficients of the prediction function in each prediction signal forming section are set in advance to coefficients of the prediction function that are adapted to different representative images.

The predicted signal holding units 34 to 36 correspond to the respective predicted signal forming units 31 to 3.
3, and supplies the held prediction signals to difference processing units 37 to 39.

On the other hand, the difference processing units 37 to 39 each include the addition processing unit 14.
．． 22, the difference processing units 37 to 39 calculate the difference between the restored signal and the predicted signal, and the difference value between both signals is sent to the subsequent discrimination processing unit 40 as a prediction error signal. Supplied.

Then, in the discrimination processing section 40, the difference processing sections 37 to 39
Using the prediction error signal indicating the prediction error value calculated in , the prediction signal forming unit that outputs the prediction error signal indicating the prediction error value with the minimum absolute value is determined, and the determined signal is sent to the prediction processing unit 41 supply to.

The prediction processing section 41 shown in FIG. 3 corresponds to the prediction processing section 16.24 of FIG. 1 and FIG. Forming part 3
2. Among the three types of prediction signals supplied from the third prediction signal forming section 33, the prediction signal with the smallest absolute value of the prediction error is selected according to the discrimination signal supplied from the discrimination processing section 40, and the selected prediction signal is selected. The predicted signal is processed by the difference processing unit 1 in FIG.
1 or the addition processing unit 14 in FIG. 1 and the addition processing unit 22 in FIG.

In the embodiment shown in FIG. 3, a case has been described in which three types of predicted signal setting sections are used, but the number of predicted signal setting sections may be increased or decreased as necessary.

FIG. 4 is a diagram showing another embodiment of the optimal prediction function calculation processing section and prediction processing section in FIGS. 1 and 2.

In FIG. 4, 51 is a coefficient storage unit that stores coefficients of a prediction function, 52 is a prediction signal calculation processing unit, 53 is a prediction signal storage unit, 54 is a discrimination processing unit, and 55 is a prediction processing unit.

The operation processing shown in FIG. 4 will be explained below.

In FIG. 4, a coefficient storage unit 51 stores coefficients of a plurality of types of prediction functions, and a prediction signal calculation processing unit 52 reads out coefficients of a plurality of types of prediction functions stored in the coefficient storage unit 51. A plurality of types of predicted signals are calculated using the restored signal supplied from the addition processing unit 14.22, and the calculated plural types of predicted signals are stored in the predicted signal storage unit 53.

Incidentally, the coefficients of the plurality of types of prediction functions stored in the coefficient storage section 51 are adapted to a plurality of types of representative images, respectively, and are stored in a memory table or the like.

The predicted signal storage unit 53 stores and holds a plurality of types of predicted signals supplied from the predicted signal calculation processing unit 52, and supplies the held predicted signals to the discrimination processing unit 54.

On the other hand, the discrimination processing section 54 is supplied with the restoration signal supplied from the addition processing section 14.22.
4, the plural types of prediction signals stored in the prediction signal storage section 53 are read out, and the difference between the restored signal supplied from the addition processing section 14.22 and the prediction signal, that is, the absolute value of the prediction error is the smallest. The coefficients of the prediction function are determined and a determined signal is supplied to the prediction processing section 55.

The prediction processing unit 55 shown in FIG. 4 corresponds to the prediction processing 16.24 in FIGS.
5 selects the prediction signal with the smallest absolute value of the prediction error according to the discrimination signal supplied from the discrimination processing section 54 among the plurality of types of prediction signals supplied from the prediction signal calculation processing section 52, and calculates the selected prediction. The signal is processed by the difference processing unit 1 in Fig. 1.
1 or the addition processing unit 14 in FIG. 1 and the addition processing unit 22 in FIG.

As described above, in the embodiment shown in FIG. 4, by using a memory table in which the coefficients of a plurality of prediction functions are stored in the coefficient storage unit 51, a small-scale hardware can be used as shown in FIG. It is possible to set coefficients corresponding to more prediction functions than in the illustrated embodiment.

[Effects of the Invention] As described above, according to the present invention, by focusing on the correlation of partial image regions, for example, the region where the correlation between adjacent pixels is high, the pixels that are currently being encoded are By finding the coefficients of the optimal prediction function for the previous pixel and using the prediction function of these coefficients to process the current pixel, it becomes possible to always perform processing using a prediction function that is suitable for each image and pixel.
It becomes possible to improve the compression effect.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the principle of predictive encoding processing to which the present invention is applied, as an embodiment of the present invention. FIG. 2 is a diagram showing the principle of decoding processing (band expansion processing) corresponding to the predictive encoding processing shown in FIG. 1. FIG. 3 is a diagram showing an embodiment of the optimal prediction function calculation processing section and prediction processing section in FIGS. 1 and 2. FIG. 4 is a diagram showing another embodiment of the optimal prediction function calculation processing section and prediction processing section in FIGS. 1 and 2. FIG. 5 is a diagram showing the principle of conventional predictive encoding processing. FIG. 6 is a diagram showing the appearance frequency 115 of prediction errors. 11... Difference processing section, 12... Nonlinear quantization processing section, 13... Encoding processing section, 14.22... Addition processing section, 15.23... Optimal prediction function calculation processing section , 16.2
4... Prediction processing section, 21... Decoding processing section.

Claims (1)

[Claims] A method for performing band compression by predictively encoding an image signal, in which the current pixel is calculated using a prediction function of coefficients that minimizes the absolute value of the prediction error with respect to the previously encoded previous pixel. A band compression method characterized by obtaining predicted values.