JPH0965334A - Image coder and image decoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain information compression with higher efficiency by allowing an inverse quantization section to use an optional value from a minimum value to a medium in a quantization step for an inverse quantization value so as to suppress a quantization error. SOLUTION: An input image signal is given to a conversion section 101, in which signal conversion such as discrete cosine transformation or sub band division is applied and quantized by a quantization section 102. A coefficient signal quantized is divided into blocks by a block processing section 103 and subject to variable length coding by a variable length coding section 104 in the unit of blocks. The coded code word is given to a variable length decoding section 105, in which the code is variable-length-coded. Then inverse processing of block processing is applied by an inverse block processing section 106 and inverse-quantization is applied by an inverse quantization section 107. Then an inverse transformation section 108 applies inverse transformation to the discrete cosine transformation or the sub band division and the result is outputted as a decoded image signal. Furthermore, an inverse quantization section 107 uses a value in the quantization step closer to 0 from a median for the inverse quantization value to reduce the quantization error.

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 for subjecting an input image signal to signal compression, quantization, blocking, and variable length coding for compression coding, and an image decoding for carrying out the decoding processing. It relates to the device.

[0002]

2. Description of the Related Art In image coding, the amount of information is reduced by reducing the redundancy of images, and variable length coding is performed to compress and decompress image data while visually suppressing deterioration of image quality. Many techniques for applying have been proposed.

FIG. 6 shows an example of a conventional image coding apparatus and image decoding apparatus. In FIG. 6, 601 is a conversion unit that performs signal conversion of an input image signal, 602 is a quantization unit that performs quantization, 603 is a blocking unit that divides the signal into blocks, and 604 is a variable-length code for the blocked signal. A variable-length coding unit for coding, and the image coding apparatus includes
It consists of 604.

Reference numeral 605 denotes a variable length decoding unit for performing variable length decoding on a codeword, 606 is an inverse blocking unit for performing inverse processing of the blocking unit 603, 607 is an inverse quantization unit for performing inverse quantization, 608. Is an inverse transform unit that performs the inverse transform of the signal transform performed by the transform unit 601, and the image decoding device is composed of 605 to 608.

The signal flow of the image coding apparatus will be described.
The input image signal to be encoded is subjected to signal conversion such as discrete cosine conversion or sub-band division by the conversion unit 601,
It is quantized in the quantizer 602. The quantized coefficient signal is divided into blocks by the blocking unit 603, and variable length coding is performed by the variable length coding unit 604 on a block-by-block basis. The output of the variable length coding unit 604 becomes the code word after information compression.

Next, the signal flow of the image decoding apparatus will be described. The code word encoded by the image encoding device is input to the variable length decoding unit 605 and variable length decoded. After that, the inverse blocking unit 606 performs inverse processing of blocking,
The inverse quantization unit 607 performs inverse quantization. The inversely quantized signal is subjected to discrete cosine inverse transformation, subband reconstruction, or the like in the inverse transformation unit 608, and is output as a decoded image signal.

Below, a quantizer 602 and an inverse quantizer 6
A conventional example of the quantization / inverse quantization processing in 07 will be described.

FIG. 7 shows the relationship between the transform coefficient signal and the quantized coefficient signal in the conventional linear quantization and inverse linear quantization, Q is the quantization step width, and C ″ m is the inverse quantized value. At this time, the inverse quantized value C ″ m becomes the central value of the quantization step. However, when the transform coefficient signal is larger than −Q and less than + Q, the quantized coefficient signal is 0, and the inverse quantized value C ″
0 also becomes 0. The processing in the quantization unit 602 and the dequantization unit 607 will be specifically described using mathematical expressions.

The quantizer 602 quantizes by the equation 1.

[0010]

[Equation 1]

Here, C'represents a quantized coefficient signal, C represents a transform coefficient signal, and Q represents a quantization step width. However, {} represents rounding down after the decimal point. Similarly, in the following formula, {}
Means truncation after the decimal point.

In addition, the inverse quantizer 607 performs inverse quantization according to equations (2), (3) and (4).

[0013]

[Equation 2]

[0014]

(Equation 3)

[0015]

(Equation 4)

Here, C ″ represents an inverse quantized value. By the processing of the above-mentioned equations 1 to 4, the quantization / inverse quantization as shown in FIG. 7 can be realized.

Next, a conventional example of the scanning process of the coefficient signal in the blocking unit 603 will be described. As a conventional scan process, there is a scan process described in Japanese Patent Laid-Open No. 3-165191 which utilizes correlation in the horizontal and vertical directions. FIG. 8 shows an example of conventional scan processing of a blocked coefficient signal.

FIG. 9 shows an example of the configuration of the conversion unit 601 which divides the input image signal into four subbands. 901, 90
5, 907 are low-pass filter units, and 902, 906, 9
08 is a high-pass filter unit, 903, 904, 909,
Reference numerals 910, 911, and 912 are thinning units. The input image signal is input to the low-pass filter unit 901 and the high-pass filter unit 902, and is filtered in the horizontal direction. After that, the thinning portions 903 and 904 thin the horizontal direction to 1/2. Similarly, thinning units 903 and 904
Is filtered in the vertical direction by the filter units 905 to 908, and is vertically thinned by the thinning units 909 to 912. With the configuration shown in FIG. 9, the input image signal can be divided into four subbands.

Here, LH is ½ in the horizontal direction through a low-pass filter (hereinafter referred to as LPF) in the horizontal direction.
The coefficient signal is obtained by performing thinning-out, and further vertically thinning out through a high-pass filter (hereinafter referred to as HPF). That is, as compared with the input image signal, the horizontal and vertical directions are decimated by 1/2, so that the number of pixels of each coefficient signal becomes ¼, and the coefficient signals for four bands are combined to form the original input image signal. It has the same number of pixels. There are the following four types of output coefficient signals after subband division.

LL ... Horizontal direction (L), Vertical direction (L) LH .... Horizontal direction (L), Vertical direction (H) HL ..... Horizontal direction (H), Vertical direction (L) HH ... ... Horizontal direction (H), vertical direction (H) (L: low range, H: high range) Since LH passes HPF in the vertical direction, correlation remains only in the horizontal direction. On the other hand, HL passes HPF in the horizontal direction, so
Correlation remains only in the vertical direction. Therefore, as shown in FIG. 8, coefficient signals having a correlation in the horizontal direction are blocked in the horizontal direction for horizontal scanning, and coefficient signals having a correlation in the vertical direction are blocked in the vertical direction for vertical scanning. As a result, the compression efficiency by the run length encoding is improved.

Further, a conventional example of the variable length coding unit 604 will be described. Huffman coding and run-length coding are widely known as variable-length coding methods effective for information compression of image signals. These encodings refer to the Huffman table created from the occurrence probabilities of the combinations of the number of zero coefficients (hereinafter, referred to as “run”) and the value of the non-zero coefficient that follows (hereinafter, referred to as “level”) Information compression is performed by assigning short codewords to high-probability events and long codewords to low-occurrence events. An example of performing variable length coding by switching a plurality of Huffman tables is the variable length coding method described in Japanese Patent Laid-Open No. 343577/1992. This is a method of switching the Huffman table according to the type of input signal, and has a configuration as shown in FIG.

Reference numeral 1101 is a run-length encoding unit for calculating the run length, and 1102 is a Huffman encoding unit for outputting the Huffman code word corresponding to the combination of the run and the level output from the run-length encoding unit 1101. . At this time, the Huffman coding unit 1102 has a plurality of Huffman tables, and switches the Huffman tables based on the type information of the input signal for coding. Thus, by using a plurality of Huffman tables suitable for the type of input signal, efficient information compression becomes possible.

[0023]

However, the above-mentioned conventional example has the following problems.

Conventional quantizer 602 and inverse quantizer 607
7, the central value of the quantization step is set to the inverse quantized value C ″ m. At this time, if the appearance probability of the transform coefficient signal is uniform as shown in FIG. The quantization error is minimized when the central value of the step is the inverse quantized value, but the coefficient signal in the high frequency band of the discrete cosine transform or subband division has the appearance probability as shown in FIG. When the quantization step width is Q, the distribution of appearance probabilities at arbitrary quantization steps in the region where the transform coefficient signal is positive is as shown in Fig. 10C. Is not uniform, and the probability of appearance is higher for a signal with a smaller absolute value of the transform coefficient signal, and conversely, the probability of appearance is lower for a signal with a larger absolute value of the transform coefficient signal.
In this case, the inverse quantized value C ″ m that minimizes the quantization error.
Is not the center value of the quantization step but the intermediate value of the center value of the quantization step from the minimum value of the quantization step.
Therefore, the conventional quantization method has a large quantization error.

Next, in the conventional blocking unit 603, the compression efficiency is improved by scanning by utilizing the correlation in the horizontal and vertical directions as shown in FIG. However, in this scanning method, since the coefficient signals are not adjacent to each other at the end of the block, the correlation between the coefficient signals is low, which hinders the coding efficiency.

Further, in the conventional variable length coding unit 604, variable length coding is performed by the configuration of FIG. 11 as described above. At this time, since the Huffman table is switched based on the type information of the input signal, when the characteristic of the input coefficient signal changes significantly, the Huffman table is not optimally encoded. That is, in the conventional variable length coding unit as shown in FIG. 11, it is not possible to dynamically respond to the characteristic change of the input signal coefficient, and it is inevitable to reduce the coding efficiency.

In view of the problems of the prior art, it is an object of the present invention to provide an image coding apparatus and an image decoding apparatus capable of suppressing quantization error and performing more efficient information compression.

[0028]

In order to solve the above problems, the present invention has the following configurations and features.

(1) A converter for converting an image input signal,
An image code including a quantizer for quantizing the transformed coefficient signal, a blocking unit for blocking the quantized signal, and a variable length coding unit for variable length coding the blocked signal. Device, variable-length decoding unit for variable-length decoding codewords, deblocking unit for performing inverse processing of blocking, dequantization unit for performing dequantization, and inverse transformation of transformation unit. In the image decoding device including the inverse transform unit, the inverse quantization unit sets an arbitrary value from the minimum value of the quantization step to the central value as the inverse quantization value.

(2) a converter for converting an image input signal,
An image code including a quantizer for quantizing the transformed coefficient signal, a blocking unit for blocking the quantized signal, and a variable length coding unit for variable length coding the blocked signal. Device, variable-length decoding unit for variable-length decoding codewords, deblocking unit for performing inverse processing of blocking, dequantization unit for performing dequantization, and inverse transformation of transformation unit. In the image decoding apparatus including the inverse transform unit, the inverse quantization unit sets an arbitrary value from the minimum value of the quantization step to the central value as an inverse quantization value, and further sets the inverse quantization value as an integer.

(3) a converter for converting an image input signal,
An image code including a quantizer for quantizing the transformed coefficient signal, a blocking unit for blocking the quantized signal, and a variable length coding unit for variable length coding the blocked signal. Device, variable-length decoding unit for variable-length decoding codewords, deblocking unit for performing inverse processing of blocking, dequantization unit for performing dequantization, and inverse transformation of transformation unit. In an image decoding apparatus including an inverse transform unit, a blocking unit scans horizontally for signal components having a correlation in the horizontal direction and vertically for signal components having a correlation in the vertical direction, and ends the block. In, a return scan is performed.

(4) a conversion unit for converting an image input signal,
An image code including a quantizer for quantizing the transformed coefficient signal, a blocking unit for blocking the quantized signal, and a variable length coding unit for variable length coding the blocked signal. Device, variable-length decoding unit for variable-length decoding codewords, deblocking unit for performing inverse processing of blocking, dequantization unit for performing dequantization, and inverse transformation of transformation unit. In the image decoding device including the inverse transform unit, the variable length coding unit includes a plurality of table units, and the table units are switched according to the ratio of zero coefficients in the block to perform variable length coding. A plurality of table units are provided, and variable length decoding is performed by switching to a table unit corresponding to the encoded table.

[0033]

The present invention has the following functions due to the above-mentioned structure.

(1) By setting an arbitrary value from the minimum value of the quantization step to the central value as the inverse quantization value, the quantization error is reduced and the amount of information after encoding is not increased. A decoded image with better image quality can be obtained.

(2) It is not necessary to treat the coefficient signal as a real number by setting an arbitrary value from the minimum value of the quantization step to the central value as the inverse quantization value and limiting the representative value to an integer. In addition, there is no quantization error that occurs when the coefficient signal is converted from a real number to an integer. Therefore, it is possible to obtain a decoded image with better image quality without increasing the amount of information after encoding.

(3) A signal component having a correlation in the horizontal direction is scanned in the horizontal direction, a signal component having a correlation in the vertical direction is scanned in the vertical direction, and a folding scan is performed at the end of the block, thereby performing block scanning. Since the scanning can be performed while maintaining the correlation between the coefficient signals at the ends of, the compression efficiency of the information amount after encoding can be improved.

(4) The variable length coding unit has a plurality of tables, and the tables are switched according to the ratio of zero coefficients in the block to perform the variable length coding. The variable length decoding unit includes a plurality of tables, and switches to a table corresponding to the encoded table to perform variable length decoding. By this means, it is possible to dynamically select a Huffman table suitable for the characteristics of the coefficient signal to be encoded, so that it is possible to improve the compression efficiency of the information amount after encoding.

[0038]

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

(Embodiment 1) FIG. 1 is a diagram showing the configurations of an image coding apparatus and an image decoding apparatus according to the first embodiment of the present invention. In FIG. 1, 101 is a conversion unit that performs signal conversion of an input image signal, 102 is a quantization unit that performs quantization,
Reference numeral 103 denotes a blocking unit that divides a signal into blocks, and 10
Reference numeral 4 denotes a variable-length coding unit that performs variable-length coding on the blocked signal, and the image coding apparatus includes 101 to 104. Further, 105 is a variable length decoding unit that performs variable length decoding of the codeword, 106 is an inverse blocking unit that performs the inverse processing of the blocking unit 103, 107 is an inverse quantization unit that performs inverse quantization, and 108 is a conversion unit. The image decoding apparatus 105 is an inverse conversion unit that performs the inverse conversion of the signal conversion performed by 101.
~ 108.

The signal flow of the image coding apparatus will be described.
The input image signal to be encoded is subjected to signal conversion such as discrete cosine conversion or subband division by the conversion unit 101,
It is quantized by the quantizer 102. The quantized coefficient signal is divided into blocks in the blocking unit 103,
The variable length coding unit 104 performs variable length coding on a block-by-block basis. The output of the variable length coding unit 104 becomes a code word.

Next, the signal flow of the image decoding apparatus will be described. The code word encoded by the image encoding device is input to the variable length decoding unit 105 and variable length decoded. afterwards,
The inverse blocking unit 106 performs inverse processing of blocking, and the inverse quantization unit 107 performs inverse quantization. The inversely quantized signal is inversely transformed by signal transformation such as discrete cosine transformation or subband division in the inverse transformation unit 108 and output as a decoded image signal.

The processing in the quantizing unit 102 and the inverse quantizing unit 107 in the case of linear quantization with the quantization step width as Q will be described using a specific example.

The quantizing unit 102 quantizes by the equation 5.

[0044]

(Equation 5)

Here, C'represents a quantized coefficient signal, C represents a transform coefficient signal, and Q represents a quantization step width.

Further, the inverse quantization unit 107 performs inverse quantization by the equations (6), (7) and (8).

[0047]

(Equation 6)

[0048]

(Equation 7)

[0049]

(Equation 8)

Here, C ″ represents an inverse quantized value, and n> 2.
And In this embodiment, a case where n = 3 will be described.
When the transform coefficient signal C is a positive number, the inverse quantization value C ″ becomes a value obtained by adding Q / 3 to the minimum value of the quantization step. On the other hand, when the transform coefficient signal C is a negative number, the inverse quantization value C ″ is The quantization value C ″ is a value obtained by subtracting Q / 3 from the maximum value of the quantization step. FIG. 2 shows the relationship between the transform coefficient signal and the quantized coefficient signal in the above quantization / inverse quantization processing. The inverse quantization value takes a value closer to 0 than the center value of the quantization step.

In the transform unit 101, the discrete cosine transform,
Alternatively, the coefficient signal in the high frequency band after sub-band division has an appearance probability as shown in FIG. Therefore, the distribution of the appearance probabilities at arbitrary quantization steps in the region where the quantized coefficient signal is positive is as shown in FIG.

As shown in FIG. 3B, since the quantization coefficient signal has a probability of taking a value closer to 0 than the center value of the quantization step, the dequantized value C ″ is larger than the center value of the quantization step. By taking a value close to 0, the quantization error is reduced.

As described above, by setting the inverse quantization value to a value closer to 0 than the center value of the quantization step, the quantization error is reduced and a better decoded image can be obtained.

Although the case of n = 3 has been described in the first embodiment, n is not limited to 3.

The signal conversion performed by the conversion unit 101 in the first embodiment is not limited to discrete cosine conversion or subband division.

(Embodiment 2) In the above-mentioned Embodiment 1, the explanation was made by using the equations 5 to 8. However, this is not limited to the equations 5 to 8, and the embodiments described below are also applied. The same effect as 1 can be obtained. The second embodiment of the present invention will be described below.

The configurations of the image coding apparatus and the image decoding apparatus in the second embodiment are the same as those in the first embodiment described above.

The quantizing unit 102 quantizes by the equations (9) to (11).

[0059]

[Equation 9]

[0060]

(Equation 10)

[0061]

[Equation 11]

Here, C ′ represents a quantized coefficient signal, C represents a transform coefficient signal, and Q represents a quantization step width, and n> 2.

Inverse quantization section 107 performs inverse quantization according to equation (12).

[0064]

(Equation 12)

Here, C ″ represents an inverse quantized value.
Even in the quantization / inverse quantization processing by the equation 12, the inverse quantization value takes a value closer to 0 than the central value of the quantization step,
The same effect as that of the first embodiment described above can be obtained.

Although the case of n = 3 has been described in the first embodiment, n is not limited to 3.

The signal conversion performed by the conversion unit 101 in the first embodiment is not limited to discrete cosine conversion or subband division.

(Third Embodiment) Next, a third embodiment will be described.

The configurations of the image coding apparatus and the image decoding apparatus according to the present embodiment are the same as the configurations of the image coding apparatus and the image decoding apparatus according to the first embodiment described above.

The quantizing unit 102 quantizes by the equation 5. Here, C ′ represents a quantized coefficient signal, C represents a transform coefficient signal, and Q represents a quantization step width.

Further, the inverse quantization unit 107 performs inverse quantization by the equations 6, 7, and 8. Here, C ″ represents an inverse quantized value, and n> 2.

At this time, Q / n is such that the value of Q / n becomes an integer.
By using n and n, C ″ is always an integer. For example, when n = 3 and the quantized value Q is a multiple of 3, the inverse quantized value C ″ is an integer. Therefore, the quantization error that occurs when the inverse quantized value C ″ is rounded down to the integer and the integer is reduced, and a decoded image with better image quality can be obtained.

Fourth Embodiment Next, a fourth embodiment will be described.

The configurations of the image coding apparatus and the image decoding apparatus of the present embodiment are the same as the configurations of the image coding apparatus and the image decoding apparatus of the first embodiment described above.

Here, in the conversion unit 101, when the input image signal is divided into four subbands and encoded, LL, LH, HL, and HL, as in the prior art example described above.
The frequency is divided into four HH bands. LH has a correlation in the horizontal direction, and HL has a correlation in the vertical direction. The configuration example of the conversion unit 101 at this time is as shown in FIG. 9 similarly to the conventional technology example. As described above, the scanning process of the coefficient signal in the blocking unit 103 when the coefficient signal having the correlation in the horizontal direction or the vertical direction is encoded will be described in detail below.

FIG. 4 shows an embodiment of the scanning process of the block-shaped coefficient signal. Coefficient signals having a correlation in the horizontal direction are blocked in the horizontal direction for horizontal scanning, and coefficient signals having a correlation in the vertical direction are blocked in the vertical direction for vertical scanning. At this time, the end of the block is folded back and scanned as shown in FIG. 4, so that the first coefficient signal to the last coefficient signal of the block can be scanned without losing the correlation of adjacent coefficient signals. Therefore, since more zero coefficient signals continue, the compression efficiency by the variable length coding is improved without degrading the image quality.

In the fourth embodiment, the conversion unit 10
1 has LL, LH, HL, HH with the configuration shown in FIG.
Although the description has been made using the case where the frequency is divided into four bands, it is clear that the same effect can be obtained by applying the present invention by dividing the frequency into smaller subbands.

(Fifth Embodiment) Next, a fifth embodiment will be described.

The configurations of the image coding apparatus and the image decoding apparatus according to the present invention are the same as the configurations of the image coding apparatus and the image decoding apparatus according to the first embodiment described above.

Variable length coding unit 104 / Variable length decoding unit 1
The operation of 05 will be described below. FIG. 5 is a diagram showing the configuration of the variable length coding unit 104 of the present invention in more detail. 501
Is a run length encoding unit for calculating the run and level, and 502 is a Huffman encoding unit for outputting the Huffman code word corresponding to the combination of the run and level output from the run length encoding unit 501. Further, reference numeral 503 is a zero ratio calculator that calculates the ratio of the zero coefficient of the input signal and outputs it as zero ratio information.

The input signal is converted into run level information by the run length encoding unit 501. Zero ratio calculator 50
In step 3, the zero coefficient ratio of the block to be encoded is calculated and the zero ratio information is output. The Huffman coding unit 502 has a plurality of Huffman tables created by sample data having different zero coefficient ratios, and dynamically selects and changes a plurality of Huffman tables using the zero coefficient ratio of the input signal as a threshold value. Performs long encoding. Further, the identification number of the Huffman table selected at this time is output as selection table information.

The variable length decoding unit 105 selects the Huffman table based on the selection table information and performs decoding.

As described above, by dynamically selecting the optimum Huffman table based on the ratio of zero coefficients in the block, it is possible to avoid a decrease in coding efficiency when the characteristics of the input image signal change. Therefore, more general-purpose information compression becomes possible.

[0084]

As described above, according to the present invention, a good decoded image can be obtained by setting an arbitrary value from the minimum value of the quantization step to the central value as the inverse quantization value. By performing quantization / inverse quantization such that the inverse quantization value is an integer at times, a better decoded image can be obtained.

Next, the coefficient signal having the correlation in the horizontal direction is scanned in the horizontal direction, and the coefficient signal having the correlation in the vertical direction is scanned in the vertical direction, and the aliasing scan is performed at the end of the block. From the coefficient signal of 1 to the last coefficient signal, efficient coding is possible without losing the correlation between adjacent coefficient signals.

Furthermore, by selecting the Huffman table based on the ratio of the zero coefficient signal in the block, the Huffman table suitable for the characteristics of the input pixel signal can be coded, which is more versatile and efficient. It is possible to compress various information.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an image encoding device and an image decoding device according to first to fourth embodiments of the present invention.

FIG. 2 is a diagram showing a relationship between a transform coefficient signal and a quantized coefficient signal according to the first embodiment of the present invention.

FIG. 3 is a diagram showing a relationship between transform coefficient signals and appearance probabilities.

FIG. 4 is a diagram showing a block scanning method according to a fourth embodiment of the present invention.

FIG. 5 is a configuration diagram of a variable length coding unit according to a fifth embodiment of the present invention.

FIG. 6 is a configuration diagram of a conventional image encoding device and image decoding device.

FIG. 7 is a diagram showing a relationship between a transform coefficient signal and a quantized coefficient signal.

FIG. 8 is a diagram similarly showing a block scanning method.

FIG. 9 is a block diagram of a conversion unit that similarly divides the subband into four

FIG. 10 is a diagram showing a relationship between a quantized coefficient signal and an appearance probability.

FIG. 11 is a block diagram of a variable length coding unit.

[Explanation of symbols]

101,601 Transform unit 102,602 Quantization unit 103,604 Blocking unit 104,604 Variable length coding unit 105,605 Variable length decoding unit 106,606 Deblocking unit 107,607 Dequantization unit 108,608 Inverse transform unit 501, 1101 Run length coding unit 502, 1102 Huffman coding unit 503 Zero ratio calculation unit 901, 905, 907 Low pass filter unit 902, 906, 908 High pass filter unit 903, 904, 909, 910, 911, 912 Thinning section

Claims (5)

[Claims] 1. An input image signal is converted into a coefficient signal, the coefficient signal is quantized, the quantized coefficient signal is blocked, and a code word obtained by variable-length coding the blocked coefficient signal is decoded. A variable-length decoding unit for variable-length decoding a codeword, an inverse-blocking unit for inverse-blocking a decoded coefficient signal, and an inverse-quantization for the inverse-blocked coefficient signal. An inverse quantization unit for performing quantization, and an inverse transformation unit for performing the inverse transformation of the transformation unit on the dequantized coefficient signal, wherein the dequantization unit performs the quantization from the minimum value of the quantization step. An image decoding apparatus characterized in that an arbitrary value up to the central value of a step is set as an inverse quantized value. 2. The image decoding apparatus according to claim 1, wherein the inverse quantization unit sets the inverse quantization value to an integer. 3. A conversion unit for converting an input image signal into a coefficient signal, a quantization unit for quantizing the coefficient signal, a blocking unit for blocking the quantized coefficient signal, and a blocking unit. A variable length coding unit for variable length coding the coefficient signal, wherein the blocking unit is arranged in a horizontal direction for a signal component having a correlation in the horizontal direction and in a vertical direction for a signal component having a correlation in the vertical direction. An image encoding device characterized in that a scan is performed on a line and a return scan is performed at the end of the block. 4. A conversion unit for converting an input image signal into a coefficient signal, a quantization unit for quantizing the coefficient signal, a blocking unit for blocking the quantized coefficient signal, and a blocking unit. A variable-length coding unit for variable-length coding a coefficient signal, wherein the variable-length coding unit includes a plurality of tables, and the table units are switched according to the ratio of zero coefficients in the block to perform the variable-length coding. An image encoding device characterized by: 5. An input image signal is converted into a coefficient signal, the coefficient signal is quantized, the quantized coefficient signal is blocked, and the blocked coefficient signal is represented by a percentage of zero coefficients in the block. Is a device for decoding a codeword that has been variable-length coded by switching between, a variable-length decoding unit that performs variable-length decoding of the codeword, and an deblocking unit that performs deblocking of the decoded coefficient signal. And an inverse transform unit that performs inverse transform of the inverse-blocked coefficient signal and an inverse transform unit of the transform unit on the inverse-quantized coefficient signal. The image decoding apparatus, wherein the encoding unit includes a plurality of tables, and switches to a table corresponding to the encoded table to perform variable length decoding.