JPS6376687A - Adaptation type difference encoding system - Google Patents

Abstract

PURPOSE:To raise the efficiency of compression by adaptively selecting a quantization characteristic according to an estimated value as for components expressing the brightness of colors and adaptively selecting the quantization characteristic according to the estimated value and the current value of colors as for the components expressing colors in case of selecting the quantization characteristic of a differential signal every component expressing a color picture after referring to the value. CONSTITUTION:As for the components expressing the brightness of colors, the quantization characteristic is selected by referring to the component value of the brightness of a former picture element and deciding an allowable quantization error. As for the components of a chrominance signal, the quantization characteristic is selected by referring to the component value of the brightness of the picture element which is observed now besides the said value and deciding the allowable quantization error. By treating the component of the chrominance signal like this, the encoding which is excellent in the compressibility of data with considering the visual characteristic of a human being can be executed and the deterioration of the picture does not occur.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an adaptive differential encoding method for compressing color image signals.

[Conventional technology]

In recent years, in order to put TV conference systems and full-color still image transmission into practical use, the development of compression transmission methods for digital image information has become active, and differential pulse code modulation (DPCM) is a method suitable for narrowband transmission lines. is attracting attention. The DPCM method is basically a method in which the difference between an input signal and a predicted signal is quantized, encoded, and transmitted. The color system for color images is YIQ, Y-R-Y-B-Y.
, CIELAB, CIELUB, etc. For example, RG
Colors are expressed using three parameters, such as B and YIQ. In the conventional predictive coding method using DPCM, quantization and coding were performed for each of these three parameters of the color system independently of each other.

As a result of studies on human visual characteristics, it is known that when image data is quantized, quantization errors that cannot be recognized by human vision depend on the magnitude of each color parameter. Quantization refers to making a value within a certain value range representative by a specific value within that range, thereby discretizing a continuous quantity, and the difference between the true value and its representative value is the quantization error. The effect of quantization error on human visual characteristics can be determined by adding uniform noise with limited amplitude to the parameters of the color system in advance and examining the detection limit for that amplitude. It is possible to know whether the quantization error of is tolerable.

Human vision has a masking phenomenon in which it is difficult to detect amplitude errors even if they occur in image data in areas where the image changes rapidly. In other words, the allowable level of error is different between parts where the signal changes rapidly and parts where the signal does not change much and is flat, and a fairly large error is allowed in parts where the signal strength, especially the brightness, changes rapidly. Ru. Due to this visual characteristic, when quantizing the difference from the previous value in the encoding method of previous value prediction (same for subsequent value prediction), even if the quantization error is large in the part where the difference value is large, humans cannot see the error. is not recognized, therefore, nonlinear quantization that widens the quantization range width can be adopted in parts where the difference value is large. Employing nonlinear quantization can improve data compression efficiency, and many conventional differential encoding systems employ nonlinear quantization.

This nonlinear quantization characteristic can be expressed, for example, by drawing polygonal lines passing through the origin and alternating between the allowable quantization error curve (or straight line) and the difference value axis, and intersecting the difference value axis at 45 degrees, and connecting the polygonal lines with the allowable quantization Determined by the intersection with the curve.

[Problem that the invention seeks to solve]

However, these color systems are not necessarily uniform spaces considering human visual characteristics. That is, the color difference that humans perceive when a color has the same norm varies greatly depending on its position in the color system. Therefore, the conventional predictive encoding method does not perform data compression that is compatible with human visual characteristics, and therefore does not have good compression efficiency.

Therefore, the present invention takes into account the permissible quantization error of human visual characteristics for each component of the color system expressing a color image, and can further improve the compression efficiency of color image signals without deteriorating image quality. The purpose of this paper is to present an adaptive differential encoding method.

Means for Solving Problem C] The adaptive differential encoding method according to the present invention samples a color image signal to obtain a sampled signal, forms a difference signal from the sampled signal, and generates a difference signal from the sampled signal. In this method, when adjusting the quantization characteristics of the difference signal for each component expressing a color image, the component representing the brightness of the color is quantized and encoded. The quantization characteristics of the difference signal are adaptively adjusted according to the allowable quantization error set according to the predicted value of This method adjusts the quantization characteristics of the difference signal according to the allowable quantization error set according to the component value representing the difference.

[Effect]

By setting an allowable quantization error for each component expressing a color image signal using the above-mentioned method and adjusting the quantization characteristics according to the allowable quantization error, image quality deterioration due to quantization errors can be suppressed, and Quantization efficiency can be increased.

〔Example〕

−4 = In the present invention, for the component representing the brightness of the color, the allowable quantization error is determined by referring to the brightness component value of the previous pixel and the quantization characteristic is selected, and for the color signal component, the quantization characteristic is selected by referring to the brightness component value of the previous pixel. The brightness component value of the pixel of interest is also referred to to determine the permissible quantization error and select the quantization characteristic. By handling the color signal components in this manner, it is possible to perform encoding with a high data compression rate that takes human visual characteristics into account, and no image deterioration occurs.

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to the drawings, which illustrate an example of a circuit configuration for implementing the method of the present invention.

FIG. 1 shows a transmitting/receiving system of a circuit implementing the method according to the present invention in the Y, R-Y, B-Y color system. In the present invention, it is necessary to refer to the brightness signal of the pixel of interest, that is, the luminance (Y) signal, for quantization of the color signal component. -Y,B
Note that the -Y signal is at time i. The explanation will mainly be based on the luminance signal Y series, but in the R-Y signal series and B-Y signal series, elements corresponding to circuit elements of the Y signal series have R and B after the numerical code, respectively, instead of Y. Attached. In order to distinguish between the true value and the value that has been quantized and includes a quantization error, a prime (“) is attached to the symbol of the quantized value.

In the transmission system A, the adder/subtractor 10Y receives the sampled y,
+1 and the quantized value (predicted value) y1' of the previous pixel from the predictor 12Y is calculated.

The quantizer 14Y receives the difference value yi from the adder/subtractor 10Y.

+3'i''. The quantizer 14Y receives the predicted value Y!' from the predictor 12Y and quantizes the input signal with a nonlinear quantization characteristic based on the corresponding allowable quantization error. Encoding The encoder 16Y binary encodes the quantized signal (representative value) yi.1'' from the quantizer 14Y.Encoder 1
For example, in 6Y, a short code is assigned to a representative value that appears frequently, and a long code is assigned to a representative value that appears less frequently.

The quantization characteristic switching circuit 18 outputs the output y of the predictor 12Y,
l and the representative value yi- by the delay circuit 19 for one pixel
+', the R-Y signal quantizers 14R and B-
The Y signal quantizer 14B is instructed to select and set each quantization characteristic to the optimum one. That is, the larger Vi' and the larger the difference between yi-+' and y,', the more a non-linear quantization characteristic is selected that provides coarse division to such an extent that no quantization error is recognized. This reduces the number of differential representative values and increases the allocation rate of short codes.

The quantizer 14R applies the quantized signal at time i of the RY signal to the encoder 16R and the conjunction predictor 12R.

The quantizer 14B also applies the quantized signal at time i of the BY signal to the encoder 16B and the conjunction predictor 12B.

The predictors 12Y, 12R, and 12B generally include an adder that adds a previous value to the differential quantization representative value, and a delay circuit that delays the output of the adder by one pixel. becomes the output of the predictor, and the output of the delay circuit is also applied to the adder and used to restore the input difference signal.

The DPCM outputs of encoders 16Y, 16R, and 16B are sent to receiving system B via transmission lines 20Y, 20R, and 20B. In receiving system B, decoders 30Y, 30R, and 30B decode the DPCM signals on transmission lines 20Y, 20R, and 2OB, and send them to adders 32Y, 32R, and 32B. Adder 32Y
, 32R, 32B are predictors 12Y, 12 in the transmission system A.
Predicted values from predictors 34Y, 34R, 34B similar to R, 12B
' are added to the outputs of the decoders 30Y, 30R, and 30B, respectively, and each color component signal y! . ,”+ ,,+ +
Output by=°. Adders 32Y, 32R, 32B
The outputs of the respective predictors 34Y, 34R, 34
It is also applied to B and used for signal restoration in adders 32Y, 32R, and 32B.

The decoder 30Y of the receiving system B receives the output V+' of the predictor 34Y.
The decoding characteristics are selected and set to correspond to the quantization and encoding characteristics of the transmission system A. The decoding characteristic switching circuit 40 receives the predicted value yi'' from the predictor 34Y and the yi-+' from the one-pixel delay circuit 42, and switches the decoder 30R for the R-Y signal and the decoder for the B-Y signal. The decoding characteristics of 30B are set to correspond to the quantization and encoding characteristics of the transmission system A.

Since the predicted previous value on the receiving side is the same as the predicted previous value on the transmitting side, complete restoration can be achieved in the receiving system B using this algorithm. Y signal, RY signal and B-
Although the encoding and decoding algorithms are the same for the Y signal, the allowable quantization error differs for each component, so the quantizer 14
Y, 14R, 14B and decoders 30Y, 30R, 30B
Naturally, the transmission characteristics of each individual device are different.

When the Y signal value 3't-t at sampling point i-1 is relatively small and Y( at sampling point i is relatively large, considering the difference and quantization of the signal yi+ rITi+ b, i at sampling point i, In the method of determining the permissible quantization error of each component (parameter) based on the luminance signal value yi-+' of the previous pixel, the 5A method using the masking phenomenon is A dense nonlinear quantization characteristic as shown in the figure is selected.

However, if the change in luminance of the pixel of interest is larger than that of the previous pixel, a larger quantization error is allowed in the signal representing the color than originally intended, and the data compression efficiency of the R-Y and B-Y signals decreases. It gets lower. The present invention improves data compression efficiency by selecting coarser quantization characteristics as shown in FIG. 6A by also referring to the luminance signal value of the pixel of interest. Note that FIGS. 5B and 6B are enlarged views of the vicinity of the zero point in FIGS. 5A and 6A, respectively.

If the R-Y and B-Y signals are band-limited like the NTSC signal, the sample period of the Y signal is 1/
A case can be considered in which the RY and BY signals are sampled at a sampling frequency such as 2.1/4. FIG. 2 shows sampling positions at 1/2 sampling frequency. In such a case, in a configuration in which the quantization characteristic is determined by referring to the luminance signal of the previous pixel, the sample value rVi + t)y (for the sample point i) is determined using the sample value y14' of the sample point i-2. sample value r of i-2
The differential quantization characteristic with yi-1+b 7L-2 will be set.

However, in this case, they are separated by two pixels, so Y
Since the signal correlation is low, it is inappropriate to determine the quantization characteristics using )'i-z'.

When applying the present invention to this, the same i as the sample time i-'l
What is necessary is to use the value of the Y signal in . An example of this configuration is shown in FIG. In the example of FIG. 3, ry.

The reason why the processing of y precedes the processing of by by two pixels is because the value of y" at sample time t, i-2 is required for processing r,,b, at sample time i. 3, circuit elements that are the same as those in FIG. 1 are given the same reference numerals.

In the configuration of FIG. 3, the predictor 42Y of the Y signal transmission system A calculates the Y signal of the transmission system A manually by y. 2, the predicted value yi' of two pixels before is output, and the predictor 44Y in the corresponding receiving system B
also outputs ylo two pixels before. The quantization switching circuit 46 of the transmission system A corresponding to the quantization characteristic switching circuit 18 of FIG.
8 and 50, the quantized value y4-z precedes by 2 pixels
” and supplies switching signals to the quantizers 14R and 14B.The decoding switching circuit 52 of the receiving system B, which corresponds to the quantization characteristic switching circuit 40 in FIG.
The output yi' of Y and the quantized value Y+-z'-12= which precedes by two pixels are received by the delay circuits 54 and 56, and a switching signal is supplied to the decoders 30R and 30B.

When it is desired to use the quantized value of the immediately preceding pixel as a predicted value for the Y signal as shown in FIG. 1, the output y of the predictor 42Y is used.

It is only necessary to delay 1゛ by one pixel using a delay circuit.

In the circuit of FIG. 3, rIl for the R-Y and B-Y signals.
j+byj (see FIG. 2 for the relationship between 1 and j) becomes an input signal, and differential encoding with the previous value at sampling time j-1 is performed. The same goes for receiving system B.

In order to make more detailed settings for the quantization characteristics, as shown in FIG.
The quantized value yi-+' may also be input to (and 52). When Y i-zo and yi゛ are of the same size, the quantization characteristic of fine division is set according to yi-2'' and the difference value in the configuration of Fig. 3, regardless of 3'i-1'. However, when 7+-+' is, for example, extremely small and has a large difference from yi'', a coarse quantization characteristic can be adopted, and therefore, as shown in Figure 4, yi-+
It is preferable to also refer to the value of ``.In this way, R-Y, B
- Even if the sampling frequency of the Y signal is lower than the sampling frequency of the Y signal, the method of the present invention can be effectively used.

In the illustrated example, the Y signal, R-Y signal, and B-Y signal are transmitted through separate transmission paths, but in the actual configuration,
After each of Y, R-Y, and B-Y is encoded, they are converted into serial signals and sent to the transmission path, and the reception system B decomposes them into three parts and decodes them. However, since the present invention requires luminance information for decoding, the RY signal and the BY signal may be delayed by one pixel compared to the Y signal. Also, the Y signal is preceded by 1 or 2 pixels for ry and by processing, but if the Y signal processing precedes the RY signal and the BY signal by 1 or 2 pixels or more, then For example, it is possible to process one screen of the Y signal first, temporarily store the quantized data in a buffer memory, and then process the R-Y signal and B-Y signal by referring to the contents. You may go.

In the above explanation, an example was given using the Y-R-Y-B-Y color system, but the method of the present invention can also be applied to other color systems, such as YIQ,
, La*b*, Lu*v* color system, Y, respectively.
Since L and L express brightness, the optimal quantization characteristics can be set using the predicted values of Y, L, and L. Also,
In the RGB system, the G signal approximately represents brightness, so the G signal is made to correspond to the Y signal in the illustrated example, and the R signal and B
The signals may be made to correspond to the RY signal and BY signal in the illustrated example.

〔Effect of the invention〕

As described above, according to the present invention, a differential code is adaptively applied according to the permissible quantization error of human visual characteristics for each component of a color system expressing a color image signal using a simple circuit configuration. Therefore, it is possible to present an adaptive differential encoding method that can reduce image quality deterioration and increase the compression efficiency of color image signals.

[Brief explanation of the drawing]

FIG. 1 shows an example of a transmitting/receiving circuit configuration for implementing the adaptive predictive coding method according to the present invention, and FIG. 2 shows an NTSC signal with R-Y, - Diagram explaining the sample position relationship when sampling the Y signal, Part 3
The figure shows an example of a circuit configuration for differentially encoding and decoding the sample signal shown in FIG. 2 according to the present invention, FIG. Figure 5B,
FIGS. 6A and 6B are diagrams illustrating nonlinear quantization characteristics used in the present invention. 10Y, IOR, l0B--Adder/subtractor 12Y, 12
R, 12B, 42Y--Predictor 14Y, 14R. 14 B-quantizer 16Y, 16R, 16B-encoder 18.46・-switching circuit 19,48,50-
Delay circuit 20Y, 2OR, 20B-transmission line 30Y, 3
0R, 30B - Decoder 32Y, 32R, 32B - Adder 34Y, 34R, 34B. 44Y-Predictor 40.52-Switching circuit 42°54.
56...Delay circuit Fig. 5A 4 special opening UG3-76 [i87 (7) River Fig. 5B

Claims (1)

[Claims] A method in which a color image signal is sampled to obtain a sampled signal, a difference signal is formed from the sampled signal, and the difference signal is quantized and encoded. When adjusting the quantization characteristics of the signal, the quantization characteristics of the difference signal are adaptively adjusted according to the allowable quantization error set according to the predicted value of the component representing the brightness of the color. and for the component representing color, the quantum of the difference signal is adaptively adjusted according to the allowable quantization error set according to the predicted value of the component representing the brightness of the color and the component value representing the current brightness of the color. An adaptive differential encoding method characterized by adjusting the conversion characteristics.