JPH06296275A - Method and device for encoding image signal - Google Patents

Abstract

PURPOSE:To improve picture quality by calculating the statistical property of an image signal at every small area part, and setting the range of coefficient amplitude to be quantized to zero when a coefficient required for the constitu tion of an image is quantized corresponding to a calculated value. CONSTITUTION:An input image signal S1 is inputted to a statistical calculation circuit 41, and the statistical property, etc., is calculated at every small area part of a screen, and a calculated result S11 is classified by a classification circuit 42, and is accumulated in memory 43 as an index S12. While, the signal S1 is inputted to a conversion circuit 11 via memory 44 for delay, and is wavelet-converted, then, it is converted to a coefficient S2. The coefficient S2 is inputted to a quantizer 45, and is quantized by quatization step size information S6 decided by the accumulation quantity of buffer memory 14 and a dead zone decided by the index S12, then, it is converted to a quantized coefficient S3. The coefficient S3 receives Huffman encoding and encoding combining with a runlength code of zero, and is inputted to a multiplex circuit 15 via the buffer memory 14, then, it is outputted by multiplexing.

Description

Detailed Description of the Invention

[0001]

[Table of Contents] The present invention will be described in the following order. Industrial Application Conventional Technology (FIGS. 6 to 12) Problems to be Solved by the Invention (FIGS. 6 to 15) Means for Solving the Problems (FIGS. 1 to 5) Actions (FIGS. 1 to 5) ) Example (FIGS. 1-5) Effect of the invention

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal coding method and an image signal coding apparatus, and can be applied to, for example, a method in which an image signal is divided into components using a wavelet transform or the like and then encoded.

[0003]

2. Description of the Related Art Conventionally, in a so-called image signal transmission system for transmitting an image signal to a remote place such as a video conference system, or an apparatus for recording an image signal into a video tape recorder or a video disc recorder for reproduction. In order to efficiently use the transmission path and the recording medium, the correlation between the digitized image signals is used to efficiently encode the significant information to reduce the transmission information amount and the recording information amount, thereby reducing the transmission efficiency and It is designed to improve recording efficiency.

As a coding method utilizing the correlation of image signals, for example, a predictive coding method, DCT (Discrete Cosin)
An orthogonal transform coding method such as e Transform), a subband coding method, a wavelet transform method, or the like, in which an image signal is divided into a plurality of components and then quantized and transmitted, is used.

The predictive coding method is easy to implement and suitable for coding with a relatively low compression rate, but if the compression rate is increased, deterioration of image quality is easily detected. Orthogonal transform coding methods such as DCT are often used because high image quality can be obtained relatively easily at a high compression rate. However, a noticeable distortion occurs at the boundary of blocks or the degree of distortion of each block. However, there is a drawback in that the obstacles caused by the difference in the are different. Further, in the orthogonal transform coding method by DCT, the large amount of calculation becomes an obstacle in terms of hardware feasibility.

Further, the component division encoding such as the sub-band encoding method or the wavelet transform method is one in which the image signal is divided into a plurality of components and then quantized, etc., and a relatively high compression efficiency can be obtained. There are few noticeable distortions. Although there are differences in the deviceization depending on the method, Haa
There is also a method that can be configured extremely simply, such as a wavelet transform using an r function.

In the wavelet transform method, component division is recursively performed on low-order side components or low-frequency side components, and the components obtained as a result, that is, each coefficient, is encoded according to its characteristics. Yes, the filter for the division and reconstruction is as a filter using the Haar function,
As shown by combining the one-sample delay circuit, the multiplication circuit, and the addition circuit in FIGS. 6A to 6D, various types can be used and can be used properly according to the application.

An image coding apparatus and an image decoding apparatus using the wavelet transform method are constructed as shown in FIGS. 7 and 8, respectively. That is, first, in the image encoding device, the input image signal S1 is input to the conversion circuit 11 and converted into the coefficient S2. The conversion circuit 11 is composed of, for example, a two-dimensional wave conversion circuit shown in FIG. As a result, the conversion circuit 11 divides the input image signal S1 into a high-order side or high-frequency side component and a low-order side or low-frequency side component, and the resulting two components are thinned out every other sample. The wavelet transform shown in FIG. 10 is performed by recursively repeating the sampling operation for the low-order side or low-frequency side components.

The coefficient S2 thus obtained is shown in FIG.
It is quantized by the quantizer 12 shown in FIG. This quantizer 12
Then, the quantization step information S6 indicating the quantization accuracy that is determined so as to smooth the generated information amount by referring to the storage amount of the buffer memory 14 arranged in the subsequent stage, and the control circuit 35.
Address information Q1 indicating the type of the coefficient supplied from the CPU, that is, the position of the coefficient in the coefficient block, is stored in the ROM (Re
It is input to a reciprocal number generation circuit 36 configured by an ad only memory) or the like to generate a reciprocal number of the divisor. The reciprocal number Q6 of the divisor generated by the reciprocal number generation circuit 36 is input to the multiplier 37 together with the coefficient S2 output from the conversion circuit 11, and the multiplier 37
Is transmitted as the quantization coefficient S3.

The quantized coefficient S3 is input to the variable length coding circuit 13 and, for example, variable length coding in which Huffman coding and zero run length coding are combined is applied to form variable length coded data S4, which is the generated information. It is input to the multiplexing circuit 15 via the buffer memory 14 for quantity smoothing, and is multiplexed with the quantization step size information S6 here to become the output S7 of the image coding apparatus.

On the other hand, in the image decoding apparatus, the input data S21 is input to the shunt circuit 21. This shunt circuit 21
Outputs the quantized coefficient S22 and the quantized step size information S23, which are variable-length coded, from the input data S21. The variable-length coded quantized coefficient S22 is input to the variable-length decoding circuit 23 via the buffer memory 22, variable-length decoded, and quantized coefficient S2.
It becomes 4.

The quantized coefficient S24 is inversely quantized in the inverse quantizer 24 in accordance with the quantized step size information S23 to become an inverse quantized coefficient S25. This inverse quantization coefficient S2
5 is input to the inverse conversion circuit 25 and inversely converted, and this is output as the restored image signal S26 which is the output of the image decoding apparatus. The inverse transform circuit 25 is composed of, for example, the two-dimensional wavelet inverse transform circuit shown in FIG. 12, and zeros are inserted between each sample of the low-order side or low-frequency side component and the high-order side or high-frequency side component. Thus, the operation of upsampling and reconstructing the data is sequentially repeated on the higher order side or the higher range side to perform the inverse transform wavebet inverse transform shown in FIG. 10 to restore the image signal.

[0013]

By the way, although the quantization and dequantization in the above-mentioned image coding apparatus and image decoding apparatus are usually performed with linear characteristics, values near zero are not changed in variable-length coding. In order to increase the compression efficiency, that is, in order to increase the number of coefficients quantized to zero, an operation of expanding the range of coefficients quantized to zero as shown in FIG. 13 may be performed.

The range of the coefficient quantized to zero is generally called a dead zone of quantization. By expanding this width from the quantization step size, the probability that the quantized coefficient becomes zero is increased, and the variable of the latter stage is changed. Many image coding methods employ a device for increasing the compression rate in long coding.
Also in the wavelet transform, it is effective to apply the dead zone to the quantization of the coefficient on the higher order side. Particularly in the case of the wavelet transform, there are a plurality of coefficients of the same type of components, that is, coefficients having the same basis function shape on the high-order side. Therefore, in general, there is a uniform dead zone for each type of component with respect to those coefficients. Is applied.

However, the plurality of coefficients of the same kind on the higher-order side are limited in the area influencing the reconstruction of the image as shown in FIG. 14, so that the width of the dead zone is irrespective of the pattern of the area. Is uniform, it is not necessarily suitable when considering the visual image quality of the restored image.
In particular, in the case of a flat pattern and a part that changes rapidly, if the width of the dead zone is widened uniformly, it is easy to observe the image quality of the flat part of the pattern, and on the contrary, the width of the dead zone is narrowed uniformly. Then, there is a problem that the compression efficiency is lowered.

Further, in the case of performing block completion conversion using the Haar function, if the width of the dead zone is uniformly set, each coefficient at the block boundary and the discontinuity point of the lower-order side function as shown in FIG. There is a problem that the step-like distortion due to the quantized separately, that is, the fluctuation distortion of the variation amount between two pixels at the block boundary and the discontinuity point is easily observed as a specific disturbance.

The present invention has been made in consideration of the above points, and the coefficient range to be quantized to zero can be finely set according to the image to improve the visual quality of the restored image and to improve the compression efficiency. It is intended to propose an image signal encoding method and an image signal encoding device which can be improved.

[0018]

In order to solve such a problem, the present invention provides an image signal encoding method for quantizing a coefficient S2 obtained by dividing an image signal S1 into components. When calculating the statistical properties for each of the values, and when quantizing the coefficient S2 necessary for reconstructing the image of each small area part, the range of the coefficient amplitude to be quantized to zero is set according to the calculated value S11. I set it.

In the present invention, the pixels a, b, c,
It is predicted that the fourth pixel d exists on one plane where the three pixels a, b, and c exist in the three-dimensional space represented by the horizontal and vertical coordinate values and amplitude values of d on the screen. Then, the predicted value p of the amplitude value of the fourth pixel d is obtained, and the absolute value | p−d | of the difference between the predicted value p and the amplitude value d of the fourth pixel d
And refer to its absolute value | p−d |
When quantizing the coefficient S2 necessary for reconstructing the image of the small area portion near the pixels a, b, c, and d, the range of the coefficient amplitude to be quantized to zero is set.

Further, according to the present invention, in the image signal coding method in which the image signal S1 is divided into components by using the Haar function and the coefficient S2 obtained by the division is quantized, the transformation block boundary and the low-order side function When quantizing the higher-order coefficient necessary for the reconstruction of the image around the discontinuity point, the range of the coefficient amplitude to be quantized to zero is made smaller than the other higher-order coefficients.

Further, according to the present invention, in the image signal coding apparatus for quantizing the coefficient S2 obtained by dividing the image signal S1 into components, the statistical property of each small area of the image signal S1 is calculated, and the calculation is performed. When the coefficient S2 necessary for reconstructing the image of each small area part is quantized according to the value S1, the range of coefficient amplitude to be quantized to zero is set, and Haa
When the r function is used, when the high-order coefficient S2 necessary for reconstruction of the image around the discontinuity point of the transform block boundary and the low-order basis function is quantized, the coefficient amplitude to be quantized to zero The range is made smaller than the other higher-order coefficients.

[0022]

The image signal S1 is analyzed to determine the degree to which distortion is easily observed for each small area portion on the screen, and the determination result S1
When quantizing the coefficient S2 required to restore the image of the region where distortion is easily observed by 1, the range of the coefficient amplitude to be quantized to zero is reduced to reconstruct the image of the region where distortion is difficult to observe. When quantizing the coefficients required for, the range of the coefficient amplitude to be quantized to zero is expanded and the compression efficiency by variable length coding is increased to further improve the visual image quality at the time of restoration with the same information amount. Can be improved.

When a region where distortion is likely to be observed is known in advance so that conversion is performed using the Haar function, the coefficient S2 necessary for restoring the image of the region is quantized to zero. The range of the coefficient amplitude to be quantized is reduced, and when the other coefficients are quantized, the range of the coefficient amplitude to be quantized to zero is expanded, and the level of distortion in both parts is equalized to generate a unique distortion. By suppressing, as a result, the compression rate and the visual image quality at the time of decompression can be further improved.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings.

FIG. 1 in which parts corresponding to those in FIG.
Shows an image encoding device to which the present invention is applied as a whole,
The input image signal S1 is first input to the statistical calculation circuit 41 which calculates the statistical properties and the like for each small area portion of the screen. The calculation result S11 of the statistical calculation circuit 41 is classified into several stages by a classification circuit 42 composed of a ROM or the like,
Index S12 for each small-area portion indicating whether distortion is easily observed
Is stored in the memory 43 as

On the other hand, the input image signal S1 is supplied to the statistical calculation circuit 4
1 and the delay memory 44 corresponding to the time required for the processing of the classification circuit 42, and is input to the conversion circuit 11 and subjected to the wavelet conversion as in the conventional example to obtain the coefficient S2. This coefficient S2 is input to the quantizer 45, and the buffer memory 14
Quantized step size information S6 determined by the amount of storage of the data and the dead zone determined by the index S12 classified into the classes stored in the memory 43, and becomes a quantized coefficient S3.

Quantizer 45 in this image coding apparatus
2 has the configuration shown in FIG. A logical product circuit 34 for forcibly setting the coefficient to zero is added. In practice, the quantization step size information S6 indicating the quantization accuracy, the coefficient address information Q1 indicating the position of the coefficient in the coefficient block supplied from the control circuit 35, and the above-mentioned classified index S12 are stored in the ROM or the like. Is input to the dead zone value generation circuit 31 composed of Dead zone generation circuit 3
1 outputs the absolute value of the maximum coefficient value quantized to zero for each coefficient to the comparison circuit 33 as the dead zone value Q2.

On the other hand, the coefficient S2 is the multiplier 37 as in the conventional case.
Is quantized and input to the logical product circuit 34, and is input to the absolute value calculation circuit 32 on the other hand, and becomes the absolute value Q3 of the coefficient,
This is output to the comparison circuit 33. In the comparison circuit 33,
The input dead zone value Q2 is compared with the absolute value Q3 of the coefficient, and when the dead zone value Q2 is larger than the absolute value Q3, the zero quantization flag Q4 connected to the AND circuit 34 is set. In the AND circuit 34, the output Q5 of the multiplier equivalent to the output of the conventional quantizer is forcibly set to zero when the zero quantization flag Q4 is set, so that the quantization of ± 1 is performed in the conventional quantization. Part of the range of coefficient amplitudes to be quantized is also quantized to zero, and the dead zone, that is, the range of coefficient values quantized to zero, is expanded as shown in FIG.

In this way, the quantizer 45 performs the quantization and the operation of the dead zone, and outputs the quantized coefficient S3.
Is input to the variable length encoding circuit 13. In the variable length coding circuit 13, for example, variable length coding in which Huffman coding and zero run length coding are combined is performed as in the conventional case, and the variable length coding quantization coefficient S4 is used for smoothing the generated information amount. It is input to the multiplexing circuit 15 via the buffer memory 14, where it is multiplexed with the quantization step size information S6 and sent as the output S7 of the image coding apparatus.

In the case of the image coding apparatus of this embodiment, as shown in FIG. 4, the statistical calculation circuit 41 has a three-dimensional space composed of horizontal and vertical pixel coordinate values and pixel amplitude values. It is designed to calculate the plane prediction in.
Here, for a small area portion P1 on the screen made up of 4 pixels, a plane prediction value predicted that a fourth pixel is present on a plane P2 in a three-dimensional space determined by the values a, b, and c of 3 pixels. 4 pixels a, b, c, with reference to the absolute value | p−d | of the difference between p = a + b−c and the actual value d of the fourth pixel.
It is determined whether or not d is on the same plane where strain is easily observed or in the vicinity of the same plane. If the class is divided into four stages, 2-bit index data S12 may be generated and stored in the memory 13 according to the magnitude of the absolute value of the difference.

The statistical calculation circuit 41 and the classification circuit 42 for performing such calculation are constructed as shown in FIG. The input image signal S1 is input to the delay memory 51, and the pixel values a, b, c, and d are output from the memory 51. From the pixel values a, b, c, d, the plane prediction value p = a-
b + c is calculated, and the prediction error p-d is calculated and input to the absolute value calculation classification circuit 52. In practice, the absolute value calculation classification circuit 52 can be configured by a ROM or the like, and in this case, the calculation of the absolute value and the classification can be processed by the same ROM. The index data S12 classified and output is stored in the memory 13.

In practice, when the Haar transform is used, in order to suppress the dead zone of the coefficient necessary for the reconstruction of the image of the part where the block boundary equal distortion is easily observed, FIG.
ROM of the dead zone value generation circuit 31 of the quantizer 45
May be set in accordance with the coefficient address so that the dead zone value of the coefficient necessary for reconstructing the image of the block boundary is smaller than the dead zone values of other coefficients.

The data encoded in this way can be decoded by the image decoding device described above with reference to FIG. 8, and the same kind of higher-order component of the same kind obtained in component division coding such as wavelet transform can be obtained. For multiple coefficients, it is easy to set the dead zone appropriately for each coefficient depending on the characteristics of the image signal, so it is possible to improve the visual quality and uniformity of the restored image and improve the compression efficiency at the same time. You can

According to the above configuration, the plurality of coefficients of the same kind of higher-order component obtained in the component division coding such as the wavelet transform are classified according to the characteristics of the image signal.
Since it becomes easy to set the dead zone appropriately for each coefficient, it is possible to improve the visual quality of the restored image and make it uniform, and further improve the compression efficiency. Thus, it is possible to realize an image signal encoding method and an image signal encoding device that can improve the performance when high quality encoding is intended.

Further, according to the above configuration, it is possible to determine whether or not the strain is easily observed for each small area portion of the screen with a simple configuration, and thus it is possible to minimize an increase in the size of the device as a whole. . In addition, even in a coding method in which a characteristic distortion is easily observed by component division coding by the Haar function, the deterioration can be suppressed to a minimum, and it is difficult to use because of the generation of the unique distortion with a simple configuration. Dea Haa
An image coding method and an image coding apparatus using the r function can be realized.

In the above embodiment, the statistical properties of each small area of the image signal are calculated in a three-dimensional space composed of horizontal and vertical pixel coordinate values and pixel amplitude values. However, the present invention is not limited to this, and the statistical properties for each small area portion of another image signal may be calculated, for example, by calculating the variance of pixels and similarly classifying. The same effect as that of the above-described embodiment can be realized.

Further, in the above-mentioned embodiment, the case where the image signal is divided into the components by the wavelet transform using the Haar function and encoded is described. The present invention is suitable for wide application to the one in which an image signal is divided into a plurality of components and encoded, such as baseband encoding.

[0038]

As described above, according to the present invention, a plurality of coefficients of the same kind of higher-order component obtained by component-division-encoding an image signal can be appropriately selected for each coefficient depending on the characteristics of the image signal. An image signal coding method and an image capable of improving the visual quality of a restored image and making it uniform by setting a dead zone and quantizing it, and at the same time improving the compression efficiency, thus realizing high-quality coding A signal encoding device can be realized.

Further, according to the present invention, it is possible to determine whether or not the strain is easily observed for each small area portion of the screen with a simple structure, and thus it is possible to minimize the size increase of the device as a whole, and Even when a coding method in which a characteristic distortion is easily observed, such as a component division coding using the Haar function, the deterioration can be minimized, so that the conventional characteristic distortion can be achieved with a simple configuration. It is possible to realize the image coding method and the image coding apparatus using the Haar function, which is difficult to use due to the occurrence of.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an image coding apparatus by an image signal coding method and an image signal coding apparatus according to the present invention.

FIG. 2 is a block diagram showing a configuration of a quantizer in the image encoding device.

FIG. 3 is a schematic diagram for explaining the expansion of a dead zone for quantization in an image encoding device.

FIG. 4 is a schematic diagram for explaining a plane prediction method as calculation of statistical properties of an image signal.

FIG. 5 is a block diagram showing a circuit configuration for generating index data by a plane prediction method.

FIG. 6 is a connection diagram showing a filter circuit for wavelet conversion using a Haar function.

FIG. 7 is a block diagram showing a conventional image encoding device.

FIG. 8 is a block diagram showing a conventional image decoding apparatus.

FIG. 9 is a connection diagram showing a configuration of a two-dimensional wavelet conversion circuit for explaining the wavelet conversion.

FIG. 10 is a schematic diagram for explaining two-dimensional wavelet transform and inverse transform using the Haar function.

FIG. 11 is a block diagram showing a configuration of a quantizer in a conventional image encoding device.

FIG. 12 is a connection diagram showing the configuration of a two-dimensional inverse wavelet transform circuit.

FIG. 13 is a schematic diagram for explaining a conventional method of expanding a dead zone during quantization.

FIG. 14 is a schematic diagram for explaining a region in which a higher-order coefficient influences in a two-dimensional inverse wavelet transform using a Haar function.

FIG. 15 is a schematic diagram for explaining a coefficient relating to a change amount between two pixels in a one-dimensional conversion using a Haar function.

[Explanation of symbols]

11 ... conversion circuit, 12, 45 ... quantizer, 13 ...
Variable length coding circuit, 14, 22 ... buffer memory, 1
5 ... Multiplexing circuit, 21 ... Shunt circuit, 23 ... Variable length decoding circuit, 24 ... Inverse quantizer, 25 ... Inverse conversion circuit, 31 ... Dead zone value generation circuit, 32 ... Absolute value Calculation circuit, 33 ... Comparison circuit, 34 ... AND circuit, 3
5 ... control circuit, 36 ... inverse number generation circuit, 37 ... multiplier, 41 ... statistical calculation circuit, 42 ... classification circuit,
43 ... Memory, 44 ... Delay memory, 51 ... Delay memory, 52 ... Absolute value calculation classification circuit.

Claims (4)

[Claims] 1. An image signal coding method for quantizing a coefficient obtained by dividing an image signal into components, wherein statistical properties of each small area portion of the image signal are calculated, and the statistical property is calculated according to the calculated value. An image signal encoding method, wherein a range of coefficient amplitude to be quantized to zero is set when quantizing the above coefficient necessary for reconstructing an image of each small area portion. 2. A fourth pixel is predicted to exist on one plane where three pixels exist in a three-dimensional space represented by horizontal and vertical coordinate values and amplitude values of the pixel on the screen. The predicted value of the amplitude value of the fourth pixel is calculated, the absolute value of the difference between the predicted value and the amplitude value of the fourth pixel is calculated, and the absolute value is referred to, in the vicinity of the four pixels on the screen. The image signal code according to claim 1, wherein a range of coefficient amplitude to be quantized to zero is set when quantizing the coefficient necessary for reconstructing an image of a small area part of Method. 3. An image signal encoding method for dividing an image signal into components by using a Haar function and quantizing a coefficient resulting from the division, wherein an image around a discontinuity point of a transform block boundary and a low-order side basis function. Image signal encoding method characterized in that, when quantizing a higher-order coefficient necessary for the reconstruction of a higher-order coefficient, the range of the coefficient amplitude to be quantized to zero is made smaller than other higher-order coefficients. . 4. An image signal encoding device for quantizing a coefficient obtained by dividing an image signal into components, calculates statistical properties of each small area portion of the image signal, and according to the calculated value, When quantizing the above coefficients necessary for reconstructing the image of each small area part, set the range of the coefficient amplitude to be quantized to zero, and when using the Haar function, transform block boundaries and low order When quantizing the higher-order coefficient necessary for the reconstruction of the image around the discontinuity point of the side basis function, the range of the coefficient amplitude to be quantized to zero is made smaller than the other higher-order coefficients. An image signal encoding device characterized by the above.