JP2690205B2 - 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 an image coding apparatus which is advantageously applied to high quality image transmission.

[0002]

2. Description of the Related Art Conventionally, DCT (discreate cosine transf)
orm), for example, Ota,
Furuseki's "Comparison of Entropy Coding Efficiency of Coefficients in MC / DCT Hybrid Coding Scheme" PCSJ
There is one shown at 86. The structure of the image coding method disclosed here is shown in FIG.

[0003] Digital image signals s which is input from input terminal 200 is divided into blocks consisting of 8 × 8 pixels, is subtracted the predicted value s _p in a subtractor 30 for each block, the prediction residuals e is output. Here the predicted value s _p is obtained as follows. That is, the input signal and the image signal s ′ one frame before stored in the frame memory (FM) 34.
Is input to the motion compensator (MC) 32, where the amount of motion between frames is detected by block matching for each 8 × 8 block component, and by compensating the amount of motion |
Let s ′ that minimizes s−s ′ | be the predicted value s _p . 8x
The prediction residual e consisting of 8 pixels is subjected to a two-dimensional discrete cosine transform in the DCT unit 36, and the transform coefficient is a quantizer (Q) 38.
Is quantized to obtain a quantized value i. The quantized coded value i is output to the variable length coding unit (VLC) 40 and the inverse quantizer (Q ~ ¹ ) 42. The code i is inversely quantized in the inverse quantizer 42, and the IDCT unit 44 performs a two-dimensional inverse discrete cosine transform. Then, the converted value e 'is added to the predicted value s _p by the adder 46, a local replay value s' is obtained. Further, the local reproduction value s'is stored in the frame memory 34.
And is used to derive the predicted value of the next frame.

The quantized value i quantized by the quantizer 38 is input to the variable length coding unit 40. In the variable length coding unit 40, the quantized value i of the DCT coefficient shown above is subjected to a zigzag scan as shown in FIG. 5, and the result of the quantization becomes “0” sandwiched by the coefficients other than “0”. Number of coefficients (run length) and size of coefficient values (level) other than "0"
Is represented by one two-dimensional code (run length level) to reduce the code amount required for transmission. For example, FIG.
In such a case, it is represented by (0, 7), (3, 5), (7, 3), and the variable length code corresponding to each is transmitted from the encoding table. Here, the variable-length code is trained using a large number of images in advance, and is created by the Huffman code according to the frequency of occurrence.

[0005]

The quantizer used in the image coding apparatus must be designed to avoid overload in consideration of the dynamic range of the input. In addition, since the influence of the quantizer on the image quality depends on the step size, unless it is made small to some extent, the image quality deterioration is detected as quantization noise in the reproduced image.
Further, by using the DCT, the amplitude value of a part of the coefficients becomes larger than that of the input. Therefore, for the purpose of high-quality image transmission, in the conventional technique, the step size is reduced to some extent, and at the same time, the number of quantization bits is increased to increase the number of levels to secure a dynamic range and avoid overload.

However, in such a conventional technique, as the number of quantization bits is increased and the number of quantization levels is increased,
There has been a problem that the number of variable length code in the variable length encoder is increased and a code length (the number of bits) is long, which leads to an increase in transmission rate.

The present invention solves the above-mentioned drawbacks of the prior art, guarantees high-quality image transmission, and reduces the amount of generated code by reducing the input dynamic range to the quantizer. The purpose is to provide.

[0008]

In order to solve the above-mentioned problems, the present invention provides an image coding apparatus for orthogonally transforming and coding an input image signal, and an orthogonal transforming means for orthogonally transforming the input image signal. And a code table having a plurality of vector codes and selecting an optimal vector code from the coefficients calculated by the orthogonal transforming means, the coefficient calculated by the orthogonal transforming means, and the vector code respectively input from the code table, The code table has a subtracting means for subtracting them, and the code table selects a vector code having a minimum difference value subtracted by the subtracting means and outputs it to the subtracting means.

[0009]

According to the present invention, the input image signal is transformed by the orthogonal transformation means, and its coefficient is sent to the code table and the subtraction means. In the code table, a vector code that minimizes the difference value subtracted by the subtraction unit is selected from the input coefficients and is output to the subtraction unit. When the vector code is input from the coefficient and code table transformed by the orthogonal transformation means, the subtraction means outputs the difference value, which is the result of subtraction, to the quantizer. When the quantizer receives the difference value, the quantizer quantizes the difference value, performs variable-length coding, and sends it to the transmission path.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of an image coding apparatus according to the present invention will be described in detail with reference to the accompanying drawings.

Referring to FIG. 1, there is shown a functional block diagram showing an embodiment of the image coding apparatus. The image coding device 10 is a coding device that performs two-dimensional discrete cosine transform on an input digital image signal, quantizes it, performs variable length coding, and sends it to a transmission line. The image encoding device has 2
The dimensional discrete cosine transform unit (DCT) 12, the codebook table 14, the subtractor 16, the quantizer (Q) 18, and the variable length coding unit (VLC) 20 are included. In the image coding apparatus according to the present embodiment, a case will be described where intraframe coding is used in which the amplitude value of the quantizer input is larger than that in interframe coding.

The input terminal 100 is an input terminal for inputting a digital image signal divided into blocks of 8 × 8 pixels, and is connected to the two-dimensional discrete cosine transform section 12. The discrete cosine transform unit 12 converts the input image signal into D
This is a calculation unit that calculates a CT to calculate a DCT coefficient s'.
The discrete cosine transform unit 12 uses the codebook table 14
And a DCT connected to the subtracter 16 and calculated to them.
Send the coefficient s'. The codebook table 14 is a DCT
It is a code selection unit that selects the optimum vector code c for input from the coefficient s'. Ie codebook table 1
In FIG. 4, n vector codes created by the pattern of generation of DCT coefficients are registered in advance by performing training using many image signals, and the power of the difference value with the input DCT coefficient s ′ is the minimum. Then, the vector code c is selected. The codebook table 14 is
It is connected to the subtractor 16 and sends it the selected vector code c.

The subtractor 16 is a subtractor which inputs the input signal s 'which is a DCT coefficient and the vector code c and subtracts the input signal s' by the vector code c. Subtractor 16
It is connected to the quantizer 18, and outputs the difference value s _p by subtracting thereto. Quantizer 18 is a quantizer for performing a quantization of the difference value s _p by a predetermined step size. Quantizer 18 is connected to the variable length coding unit 20, and outputs the difference value s _p quantized thereto. Variable length coding unit 20
Is a coding unit that codes the signal of the discrete value input from the quantizer 18 using a run length code and outputs the signal to the transmission path.

The image signal s input from the terminal 100 is
The DCT operation unit 1 is divided into blocks of 8 × 8 pixels.
2 is input. The image signal s is converted to DCT by the calculation unit 12.
The DCT coefficient obtained as a result of the calculation is input to the codebook table 14 and the subtractor 16. In the codebook table 14, the optimum vector for input is selected,
Obtaining a difference value s _p is subtracted to the input signal s' by a subtracter 16. Incidentally, the vector code power of the difference value s _p selected in the codebook table 14 is minimized, it is also transmitted to the receiving side (image decoding apparatus).

Since the vector code used in the present embodiment specifies the generation pattern of the DCT coefficient, a pattern for all coefficients of 8 × 8 pixels is not prepared, and a low-order pattern as shown in FIG. 2 is used. The code for the coefficient was set. This is because the DCT coefficient generally has a larger amplitude value of the low-order coefficient than the high-order coefficient, and thus the dynamic range of the low-order coefficient is reduced by the above method. Also,
By reducing the vector order in this way, the pattern of coefficient generation can be reduced, the number of vectors can be reduced, and the number of bits indicating the code number to be transmitted can be reduced, so the amount of transmission required to transmit the code number can be reduced. Can be reduced. .

Further, as in the image coding apparatus 25 shown in FIG. 3, by preparing k vector codebook tables and sequentially calculating the differences, the amplitude value of the quantizer input can be more surely reduced. . Further, at this time, as in the codebook tables 1, 2 to k of FIG. 3, a pattern in which a vector code does not exist in each vector code array and the pattern of the area where the vector code does not exist in each codebook table is provided. If the numbers are different, the order of each vector is reduced, the number of vectors is reduced, and the number of bits indicating each code number is reduced, thereby increasing the number of transmission bits by setting k codebook tables. Can be suppressed.

[0017] Returning to FIG. 1, the difference value s _p output from the subtracter 16 after being quantized in the quantizer 18 is input to the variable length coding unit 20. In the variable length coding unit 20, the input quantized value is zigzag scanned for each 8 × 8 pixel (see FIG. 5), and the number of “0” s sandwiched by the quantized values other than “0” and “0” are set. A combination of non-quantized quantized values is represented by one two-dimensional code, which is variable-length coded (see FIG. 6) using Huffman code and sent to the transmission path.

As described in detail above, this embodiment is
It is characterized by having a vector code created from a pattern of CT coefficients and providing a function of obtaining the difference between the DCT coefficient which is the input of the quantizer 18 and the optimum vector code.
Although the image signal s divided into blocks of 8 × 8 pixels is coded in this embodiment, the present invention is not limited to such an image signal. Further, in the present embodiment, DC is used as the orthogonal transformation to which the present invention is advantageously applied.
Although T is used, the present invention may be applied to other orthogonal transforms.

[0019]

As described above, according to the image coding apparatus of the present invention, by subtracting the optimum vector value selected from the codebook table from the quantizer input value in advance, the amplitude of the input signal to the quantizer is reduced. The value can be reduced. Therefore, even if the step size is reduced in order to perform high-quality image transmission, the number of bits can be expressed with a smaller number of bits than in the conventional quantizer, and quantization can be performed without overload. Further, by reducing the number of quantization bits in this way, the pattern of the variable length code table can be reduced, so that the run length code can be shortened and high quality image coding can be performed with a smaller number of bits than the conventional one. . Furthermore, by reducing the amplitude value of the quantizer input value, the number of coefficients that become “0” as a result of quantization increases,
In the run-length coding in which the coding is performed based on the coefficient of "0", the generated code amount is also advantageous.

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing an embodiment of an image encoding device according to the present invention,

FIG. 2 is an explanatory diagram of a vector code in the embodiment shown in FIG.

FIG. 3 is a functional block diagram showing another embodiment of the image encoding device according to the present invention and an explanatory diagram of its code table;

FIG. 4 is a block diagram showing an image encoding device in the related art;

FIG. 5 is an explanatory diagram illustrating a scan operation of a variable length coding unit.

FIG. 6 is an explanatory diagram showing encoding by a variable length encoding unit.

[Explanation of symbols]

 10, 25 image coding device 12 DCT operation unit 14 codebook table 16 subtractor 18 quantizer

Continuation of the front page (56) References JP-A-62-154987 (JP, A) JP-A-3-53665 (JP, A) JP-A-3-52385 (JP, A) JP-A-60-194686 (JP , A)

Claims (1)

(57) [Claims] 1. An orthogonal transform means for orthogonally transforming an input image signal to compute a coefficient, and a vector code having a plurality of vector codes and having a minimum difference from the coefficient computed by the orthogonal transform means. A vector table to be selected, subtraction means for subtracting the coefficient calculated by the orthogonal transformation means with the selected vector code to obtain a difference value; a quantizer for quantizing the difference value; A variable-length coding unit that performs variable-length coding on the difference value, and an image coding apparatus that outputs the selected vector code and the variable-length coded difference value, wherein the vector tables are plural. The vector codebook tables of different types are connected in multiple stages, and the difference between the vector codebook table and the coefficient calculated by the orthogonal transform means is minimized by sequentially calculating the difference. In addition to selecting a desired vector code, each vector code array of the plurality of types of vector code book tables is provided with an area where no vector code exists, and the vector code exists according to each vector code book table. An image encoding device characterized in that patterns of non-existing regions are different.