JP2872149B2 - Image coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus, and more particularly, to a code amount control circuit in an image coding apparatus for performing orthogonal transform on an input image signal and performing variable length coding.

[0002]

2. Description of the Related Art Generally, in an image coding apparatus using orthogonal transform and variable-length coding, an orthogonal transform is applied to an input image signal to impart a bias to transform coefficient values, and transform coefficient values having a high probability of occurrence are obtained. By assigning a short code word to a short code word and a long code word to a transform coefficient value with a low occurrence probability, the code amount is compressed as a whole, and encoding is performed with high efficiency. In such an image encoding apparatus, a code amount control circuit is required to encode input images having different information amounts and characteristics with a fixed image quality and a fixed code amount or less.

In a conventional code amount control circuit, one per screen is used.
A method of controlling the quantization width using one or a plurality of quantization step values has been used. In this method, when the information amount of the input image is small, the quantization step value is reduced, and the quantization coefficient value to be discarded is reduced by reducing the quantization width, and conversely, the information amount of the input image is large. In this case, a characteristic is utilized in which the number of invalid coefficient values is increased by increasing the quantization step value and increasing the quantization width to control the code amount.

[0004] In an image encoding apparatus utilizing this characteristic and performing encoding processing at a fixed code amount or less using a plurality of quantization step values for one screen, the state of an output buffer of a variable length encoding circuit is determined. If the coding process is monitored with less than the target code amount, the quantization step value is reduced, and if the target code amount is exceeded, the quantization step value is increased to reduce the generated code amount. Those using control are generally used. For example, assume that the screen is composed of a flat part such as the sky and a complex part such as a flower bed. In such a case, in this method, when the compression ratio is low, the quantization process is performed with a small quantization step value, so that sufficient coded bits are allocated to each block and uniform image quality is obtained. However, in this method, when the compression ratio is high, since the difference in local characteristics within the screen is not taken into account, the quantization process is performed using the same quantization step value in the blocks of the flat portion and the complex portion. When,
In the flat part, since the transform coefficient value has a small amplitude value, a low-frequency component having a human-sensitive visual characteristic is cut off by the quantization process, so that noise becomes remarkable and image quality deterioration is perceived.

Various prior arts related to the present invention are also known. For example, Japanese Patent Application Laid-Open No. Hei 6-113284 (hereinafter, referred to as
(Referred to as Prior Art 1) discloses an “image encoding device” that improves image compression efficiency and reduces image quality degradation. In the prior art 1, weighting quantization is performed by a quantization circuit, and the quantized DCT coefficient includes two or more spatial frequency horizontal or vertical high frequency components each having a non-zero amplitude. Then, the DCT coefficient output from the DCT circuit is selected by the selection circuit, and the quantization step in the quantization circuit is controlled.

Further, Japanese Patent Application Laid-Open No. Hei 3-3496 (hereinafter, referred to as “Japanese Patent Application Laid-open
Prior art 2), a plurality of orthogonal transforms are prepared in advance,
An “image coding method” is disclosed in which a quantizer is set depending on each of a plurality of orthogonal transforms performed on each block to improve the image quality of a decoded image. In this prior art 2, first, for example, a screen is set to 8 × 8, 16 × 1.
6, and then, for example, an autocorrelation matrix, a variance value, or a coefficient value obtained by each transform of each block,
Etc. is extracted. Next, based on this information, one optimal one is selected from a plurality of orthogonal transforms. Next, a quantization method is determined for each block or for each transform based on the information or transform of all blocks. Next, orthogonal transform is performed using the determined transform, and the transform coefficients are quantized by a method determined by each transform and coded and transmitted. As described above, by using the number of bits and the number of steps according to the conversion, it is possible to perform encoding that reflects the nature of the conversion more, and it is possible to obtain an image quality superior to the related art.

Japanese Patent Laid-Open Publication No. Hei 7-236137 (hereinafter referred to as Prior Art 3) discloses a quantization matrix in which the quantization result on the high frequency side becomes as small as possible in accordance with the number of signals to be quantized. Is disclosed, an "image coding control method" in which deterioration of image quality is suppressed. In Prior Art 3, the number of quantized signals to be quantized per block is determined by the quantized signal number determination unit from the image size and the frame rate of the input image signal. The quantization matrix selection unit selects a quantization matrix such that signals after the number of quantization signals become zero as much as possible after quantization in the zigzag scan order in the quantization unit. The input image signal is subjected to DCT by a discrete cosine unit (DCT unit) for each block, and the quantization unit quantizes the DCT result in a zigzag scan order using a selected quantization matrix, and performs quantization on the high frequency side after the number of quantized signals. The quantization of the signal is omitted and regarded as zero. The quantized signal is encoded by the variable length encoding unit and temporarily stored in the buffer memory unit. The quantization control unit monitors the buffer memory unit and generates a quantization control signal to the quantization unit.

[0008]

As described above, in the conventional image coding apparatus, the quantization process is performed with the quantization step value calculated based on the state of the output buffer. For this reason, when the compression ratio is high and the input image has different characteristics in the screen, the conventional image coding apparatus uses the same quantization step value as the block of the complicated part for the block including the flat part. When the quantization process is performed, the image quality is not perceived in a complicated part because the human visual characteristics are insensitive, but the information of the low frequency component of the transform coefficient value is not sensitive in the flat part because the visual characteristics are sensitive. Lost by the quantization process,
There is a problem that it is perceived as partial image quality deterioration.

Accordingly, an object of the present invention is to analyze the frequency distribution of transform coefficient values of a block in order to obtain a good reproduced image even in a region having a high compression ratio in which image quality deterioration is easily perceived, and to improve human visual characteristics. It is an object of the present invention to provide an image coding apparatus that reduces the number of transform coefficient values that are truncated at a sensitive flat part and makes it difficult to generate block distortion.

In the prior art 1, the quantization step of the quantization circuit is controlled so that the minimum value of the significant coefficient value becomes 1, so that the compression ratio is high and the input image has different characteristics in the screen. This is different from the case in which encoding can be performed with a fixed image quality. Prior art 2 is a technical idea of selecting both a plurality of orthogonal transforms and a quantizer depending on the orthogonal transform, and controls a quantization step value based on a result of analyzing a frequency distribution of transform coefficient values of a block. It is different from what you would do. Further, the prior art 3 is a technical idea of sequentially quantizing the number of quantized signals from the low frequency side signal in the orthogonally transformed block and quantizing the remaining high frequency side signal to zero. This is different from the method of reducing the number of transform coefficient values that are cut off in a flat part where human visual characteristics are sensitive.

[0011]

An image coding apparatus to which the present invention is applied is a block dividing means for dividing an input image signal into blocks each having a small area composed of a plurality of pixels, and an orthogonal transform for performing an orthogonal transform on the blocks. Means for quantizing a transform coefficient value which is an output result of the orthogonal transform means to obtain a quantized coefficient value.
It has a quantizing means for outputting, a quantizing control means for obtaining a quantization step value in the quantizing means, and a variable-length coding means for performing variable-length coding on the quantization coefficient value.

In the first aspect of the present invention, the quantization control means analyzes the distribution of high frequency components of the transform coefficient value of each block and outputs frequency distribution information of the transform coefficient value in each block. It comprises a frequency analysis means for, and limiter means for performing different limiter processing on the quantization step value based <br/> Zui frequency distribution information.

In a second aspect of the present invention, the quantization control means analyzes the distribution of high frequency components of the transform coefficient value of each block, and obtains the frequency distribution information of the transform coefficient value in each block. frequency distribution means for outputting, each block
Using transform coefficient values and predetermined quantization step values
Performs a quantization process and encodes the quantized coefficient value using variable-length coding.
A reference value calculation unit that outputs a reference value of the quantization step from the generated code amount when the difference is generated, a difference value calculation unit that obtains a difference value between the output code amount of the variable length coding unit and the target code amount, An adaptive limiter is provided for adaptively performing limiter processing on the sum of the difference value and the frequency distribution information and the difference value.

The quantization control means according to the first and second aspects, as the frequency analysis means, divides a frequency band of a transform coefficient value into a plurality of regions, and obtains the transform coefficient values of each region as frequency distribution information. An activity distribution unit for calculating the sum of absolute values and the sum of squares is provided.

[0015]

According to the image coding apparatus of the present invention, the frequency distribution of the transform coefficient value in each block is analyzed for each block, and a plurality of maximum quantization step values are set for each frequency distribution information. Thus, even when the compression ratio is high, a larger amount of code can be allocated to a block that is greatly affected by the quantization process. As a result, it is possible to suppress image quality deterioration of the flat portion that is easily perceived.

[0016]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows an image coding apparatus according to a first embodiment of the present invention. The illustrated image encoding apparatus includes a block division circuit 1, an orthogonal transformation circuit 2, and a frequency analysis circuit 3.
, A limiter circuit 4, a delay circuit 5, and a quantization circuit 6.
And a variable-length encoding circuit 7.

The input image signal 100 is digitized image data, and is composed of a multi-valued black-and-white image, an RGB primary color signal, luminance, and two types of color difference signals. The block dividing circuit 1 divides an input image signal 100 into blocks each including a plurality of pixels set in advance, and generates block data 10.
Outputs 1. In the orthogonal transformation circuit 2, the block data 1
01 is subjected to orthogonal transformation, and zigzag scan transformation or the like is performed so that non-zero coefficient values are continuous when variable-length coding is performed, and a transformation coefficient value 102 is output. The orthogonal transform performed by the orthogonal transform circuit 2 includes a discrete cosine transform (DCT).
Transformation) is common, but other transformations may be used.

The frequency analysis circuit 3 performs a frequency analysis of the transform coefficient value of the block. Here, the frequency analysis circuit 3
Using a maximum frequency analysis circuit for detecting the position of the maximum frequency of the non-zero transform coefficient value in the block, the frequency analysis information 10
As 3, the effective coefficient maximum frequency position is output to the limiter circuit 4. The limiter circuit 4 performs different maximum value limiting processing on the reference quantization step value 104 using the frequency analysis information 103 and outputs the result to the quantization circuit 6 as the quantization step value 105. That is, the combination of the frequency analysis circuit 3 and the limiter circuit 4 functions as a code amount control circuit (quantization control means).

On the other hand, the delay circuit 5 delays the conversion coefficient value 102 by the number of clocks necessary for the frequency analysis circuit 3 to analyze the conversion coefficient value 102, and outputs a delayed conversion coefficient value 106 to the quantization circuit 6. In the quantization circuit 6, the limiter circuit 4
Quantization of the delay transform coefficient value 106 is performed by using a quantization table (not shown) that weights the transform step value 105 and the transform coefficient value differently for each frequency.
The quantized transform coefficient value 107 is output to the variable length coding circuit 7. In the variable length coding circuit 7, the quantization conversion coefficient value 107
Is converted into variable-length data using a run-length code, a Huffman code, or the like, and a variable-length codeword 108 corresponding to the quantized conversion coefficient value 107 is output as compressed data.

As an example, the reference quantization step value 104 =
It is assumed that the quantization processing is controlled at 17 and different maximum quantization step values are set for three types of maximum frequency positions by a selection circuit (not shown). The maximum frequency position of the non-zero transform coefficient value indicated by the frequency distribution information 103 is determined by the three bands (region 1, region 2,
Divide into area 3). That is, (1) when only the low-frequency 10 components shown in the region 1 have non-zero transform coefficient values, the maximum quantization step value is set to 4, and (2) the intermediate band 2 shown in the region 2
When there are non-zero transform coefficient values in the six components, the maximum quantization step value is set to 16. (3) When there are non-zero transform coefficient values in the intermediate region 28 component shown in the area 3, the maximum quantization step value is set to 16. 32.

With this setting, when the transform coefficient value exists only in region 1, the reference quantization step value =
Although 17 is set, the quantization step value of the corresponding block is set to the maximum value = 4, so that the quantization step value = 4 is output to the quantization circuit 6. Similarly, in the case of the block corresponding to the area 2, the quantization step value = 16
Is output to the quantization circuit 6 in the block corresponding to the area 3. By performing such a process, the blocks corresponding to the region 1 and the region 2 are subjected to the quantization process with a value smaller than the preset reference quantization step value 104. More than when fixed. However, since the number of transform coefficient values is small in the blocks in the region 1 and the region 2, even if encoding is performed with a smaller quantization step value, the amount of generated codes does not increase significantly, and the quality of the reproduced image is averaged. one of image quality can be obtained.

FIG. 3 shows an image coding apparatus according to a second embodiment of the present invention. The illustrated image coding apparatus differs from that shown in FIG. 1 in that a reference quantization step value calculation circuit 8 and a difference quantization step are used instead of the limiter circuit 4 in order to perform coding with a fixed code amount or less. Value calculation circuit 9,
That is, an adder 10 and an adaptive limiter circuit 11 are provided. Components that perform the same operations as those constituting the image encoding device shown in FIG.
Hereinafter, only different points will be described.

The reference quantization step value calculation circuit 8 performs a quantization process using the absolute value of the AC component of the transform coefficient value 102 or a predetermined quantization step value, and converts the quantization coefficient value to a variable length. Code length information and the like at the time of encoding are accumulated over a certain section to estimate the amount of generated code of a block, and a reference quantization step is performed so as to be equal to or less than a predetermined target code amount 109. Calculate the value 104,
Output to the adder 10.

The difference quantization step value calculation circuit 9 accumulates the code length information of the variable length code word output from the variable length coding circuit 7, performs a difference operation with the target code amount 109, and obtains the difference quantization step value. It outputs to the adder 10 as 110. The adder 10 performs an addition process on the reference quantization step value 104 and the difference quantization step value 110, and outputs an addition quantization step value 111 to the adaptive limiter circuit 11.

The adaptive limiter circuit 11 adaptively performs a limiter process on the addition quantization step value 111 using the frequency analysis information 103 and the difference quantization step value 110, and generates a quantization step value 105 as the quantization step value 105. Output to

In any case, the combination of the frequency analysis circuit 3, the reference quantization step value calculation circuit 8, the difference quantization step value calculation circuit 9, the adder 10, and the adaptive limiter circuit 11 constitutes a code amount control circuit (quantization control means). Work as).

FIG. 4 shows an operation example of the adaptive limiter circuit 11. In the figure, the regions 1, 2, and 3 correspond to the regions 1, 2, and 3, respectively, shown in FIG. It is assumed that the difference quantization step value 110 is smaller than 2. In this case, when the effective coefficient frequency distribution information 103 indicates the area 1, the maximum value of the quantization step value is limited to 4.
Similarly, when the effective coefficient frequency distribution information 103 indicates the area 2, the maximum value is limited to 16, and when the effective coefficient frequency distribution information 103 indicates the area 3, the maximum value is limited to 32.

Next, it is assumed that the difference quantization step value 110 is 2 or more and smaller than 4. In this case, the maximum value is 5 in the area 1, the maximum value is 18 in the area 2, and the maximum value is 32 in the area 3. Further, if the difference quantization step value 110 is 1
When the number is 2 or more, the maximum value is set to 32 in each of the area 1, the area 2, and the area 3 in the same manner, and the generated code amount is suppressed assuming that the code amount is greatly exceeded.

As described above, by applying the maximum value limit to the quantization step value used for the quantization process adaptively from the frequency distribution information 103 and the difference quantization step value 110, the limited code amount can be reduced. By increasing the allocation to a flat portion where image quality deterioration is easily perceived and reducing the allocation of a complicated portion that is hardly perceived, it is possible to perform encoding with a code amount equal to or less than a certain code amount and improve image quality.

The analysis method used in the frequency analysis circuit 3 may be replaced with an activity analysis. In the activity analysis, the frequency band of the transform coefficient value is divided into a plurality of regions, the sum of absolute values or the sum of squares of the transform coefficient values in each region is obtained, and the limiter circuit 4 (FIG. 1) or the adaptive limiter circuit 11 (FIG. Output to 3).

In the embodiment shown in FIGS. 1 and 3, the maximum frequency position of the non-zero transform coefficient value is used as the maximum value switching signal of the maximum quantization step value, but by employing the activity analysis method, By controlling the limiter circuit 4 using the accumulated value of each area as a threshold value, the characteristics of the block can be analyzed more accurately than using the maximum frequency position of the transform coefficient value in the block for control. Further, the reproduced image quality can be improved.

The present invention is not limited to the above-described embodiment, and various changes and modifications can be made without departing from the spirit of the present invention.

[0034]

As described above, according to the present invention, by providing the frequency analysis circuit for analyzing the maximum frequency position of the transform coefficient value of the orthogonal transform and the limiter circuit in the code amount control circuit, the compression rate can be increased. and when the input image has different characteristics in the screen also, there is an effect that <br/> Ru can do sign-of constant quality.

Further, by providing a reference quantization step value calculation circuit, a difference quantization step value calculation circuit, and an adaptive limiter circuit in place of the above-mentioned limiter circuit, it is possible to reduce the amount of code to be allocated to a portion where human visual characteristics are insensitive. By reducing and increasing the amount of code assigned to the flat part with sensitive visual characteristics,
The quantization processing can be performed with an optimum quantization step value for each characteristic. In addition to being able to perform coding with a fixed code amount or less, there is an effect that excellent reproduction image quality can be obtained.

Further, by using the activity analysis in the frequency analysis circuit, it is possible to analyze the frequency analysis of the block of the input image in more detail, and it is possible to further improve the quality of the reproduced image.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image encoding device according to a first embodiment of the present invention.

FIG. 2 is a frequency distribution diagram of transform coefficient values used for controlling a limiter circuit of the image encoding device shown in FIG.

FIG. 3 is a block diagram illustrating a configuration of an image encoding device according to a second embodiment of the present invention.

FIG. 4 is a diagram illustrating an example of a limiter operation of an adaptive limiter circuit of the image encoding device illustrated in FIG. 3;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Block division circuit 2 Orthogonal transformation circuit 3 Frequency analysis circuit 4 Limiter circuit 5 Delay circuit 6 Quantization circuit 7 Variable length coding circuit 8 Reference quantization step value calculation circuit 9 Difference quantization step value calculation circuit 10 Adder 11 Adaptive limiter Circuit 100 Input image signal 101 Block data 102 Orthogonal transform coefficient value 103 Frequency distribution information 104 Reference quantization step value 105 Quantization step value 106 Delayed transform coefficient value 107 Quantization coefficient value 108 Variable length codeword 109 Target codeword 110 Difference quantum Quantization step value 111 Addition quantization step value

Claims (4)

(57) [Claims] A block dividing unit for dividing an input image signal into blocks each having a small area including a plurality of pixels; an orthogonal transforming unit for performing an orthogonal transform on the block; and a transform coefficient value as an output result of the orthogonal transforming unit And quantized coefficient values
, A quantization control means for obtaining a quantization step value in the quantization means, and a variable-length coding means for performing variable-length coding on the quantization coefficient value. the quantization control unit analyzes the distribution of the frequency components of the transform coefficient values for each block, a frequency analysis means for outputting a frequency distribution information of the transform coefficient values within each block, on the basis of the frequency distribution information An image encoding apparatus comprising: a limiter that performs different limiter processing on the quantization step value. 2. A block dividing means for dividing an input image signal into blocks each having a small area composed of a plurality of pixels, an orthogonal transform means for performing orthogonal transform on the block, and a transform coefficient value which is an output result of the orthogonal transform means. And quantized coefficient values
, A quantization control means for obtaining a quantization step value in the quantization means, and a variable-length coding means for performing variable-length coding on the quantization coefficient value. the quantization control unit analyzes the distribution of the frequency component of the transformation coefficient values of each block, and the frequency distribution means for outputting a frequency distribution information of the transform coefficient values within each block, Ya the conversion coefficients of each block Predefined quantization step value
Quantization is performed using
A reference value calculation unit that outputs a reference value of a quantization step from a generated code amount when encoding, a difference value calculation unit that calculates a difference value between an output code amount of the variable length coding unit and a target code amount, For the added value of the reference value and the difference value,
An image encoding apparatus comprising: an adaptive limiter that adaptively performs limiter processing using the frequency distribution information and the difference value. 3. The frequency analysis unit divides a frequency band of a transform coefficient value into a plurality of regions, and uses a sum of absolute values of transform coefficient values of each region as the frequency distribution information. 3. The image encoding device according to 1 or 2. 4. The frequency analysis unit according to claim 1, wherein the frequency band of the transform coefficient value is divided into a plurality of regions, and a sum of squares of the transform coefficient values of each region is used as the frequency distribution information. Or the image encoding device according to 2.