JPH04225693A - Encoder - Google Patents

Abstract

PURPOSE:To divide a picture signal into blocks and to efficiently encode each block in a coder in which a picture data is coded through orthogonal transformation to implement data compression. CONSTITUTION:A picture signal divided into blocks is subjected to orthogonal transform at a DCT section 101, the result is stored tentatively in a buffer memory 102, an activity calculation section 105 calculates the activity of each block and a quantity table in response to the activity is generated through a quantization factor control section 106 by an arithmetic section 107 based on a quantization table 108. Quantization of each block is implemented at a quantization section 103 by using the generated quantization table and the result is coded by a coding section 104.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoding device that compresses data by orthogonally transforming image data and encoding quantized transform coefficients.

0002

2. Description of the Related Art In recent years, various encoding methods have been studied for purposes such as band compression in processing image data, and one of them is a method using orthogonal transformation.

[0003] As a still image encoding method, focusing on the fact that images have correlation, image data is divided into blocks of N×N pixels, and data in the blocks of N×N pixels is subjected to discrete cosine transformation. There is a method of performing orthogonal transform such as (DCT) and then quantizing and encoding and compressing the obtained transform coefficients. FIG. 8 shows a block diagram of a conventional encoding device, and FIG. 9 shows characteristics of transform coefficients of discrete cosine transform. FIG. 9 shows transform coefficients when discrete cosine transform is applied to pixels of a two-dimensional block of N×N,
The shaded area is the area where the numerical value is large. The transform coefficients of blocks with many low frequency components, such as images, are concentrated in the upper left corner of the block. In the case of images, important information is concentrated in low frequency components. When such transform coefficients are quantized, the transform coefficients of low frequency components are quantized finely, and the transform coefficients of high frequency components are quantized coarsely. The encoding device has 30 components as shown in FIG.
1 is a DCT unit that performs orthogonal transformation on input image data to obtain transformation coefficients; 302 is a buffer memory that temporarily stores the transformation coefficients obtained by orthogonally transforming the DCT unit 301;
04 is a quantization table, 303 is a quantization table 304
305 is a quantization unit 303 that quantizes the transform coefficients temporarily stored in the buffer memory 302 using
This is an encoding unit that encodes the transform coefficients quantized by .

The operation of FIG. 8 will be explained below. The input image data is divided into N×N blocks. The blocked image data is orthogonally transformed in the DCT unit 301,
The obtained conversion coefficients are temporarily stored in buffer memory 302. This transform coefficient is quantized by a quantization unit 303 using a quantization table 304. The obtained result is code part 3
05. Thereafter, this process is applied to all blocks into which the image has been divided.

[0005]

In the above encoding device, the same quantization table is used for all blocks during quantization. However, since each block has a different distribution of frequency components, if all blocks are quantized using the same quantization table, encoding efficiency will deteriorate.

[0006] The present invention takes this point into consideration and attempts to provide an encoding device that efficiently encodes each block.

[0007]

[Means for Solving the Problems] In order to achieve the above-mentioned objects of the present invention, as a first invention, orthogonal transformation means for performing orthogonal transformation on image data, and coefficients for temporarily storing the results obtained by this orthogonal transformation means a storage means, an activity calculation means for calculating an activity representing the fineness of an image in the block, a table storage means for storing a first quantization table for controlling the quantization step width of each coefficient in the block; a calculation means for multiplying the first quantization table by a quantization factor to obtain a second quantization table; a quantization factor control means for obtaining the quantization factor by an activity; It has a quantization means for quantizing the coefficients in the storage means, calculates an activity for each block, obtains a quantization factor by this activity, and calculates a second quantization factor by using this quantization factor and the first quantization table. This is an encoding device that obtains a quantization table and quantizes a block using this second quantization table.

[0008] A second aspect of the invention further includes orthogonal transformation means for performing orthogonal transformation on image data, coefficient storage means for temporarily storing the results obtained by the orthogonal transformation means, and coefficient storage means for temporarily storing the results obtained by the orthogonal transformation means. activity calculation means for calculating the activity in the block; table storage means for storing a first quantization table for controlling the quantization step width of each coefficient in the block; quantization factor control means for obtaining the quantization factor according to the activity; and quantization means for quantizing the coefficients of the coefficient storage means using the second quantization table. calculate the activity for each block, obtain a quantization factor by this activity, obtain a second quantization table by this quantization factor and the first quantization table, and calculate the block by the second quantization table. This is an encoding device that performs quantization.

[0009]

[Operation] The encoding device of the present invention having the above configuration divides input image data into blocks, calculates the activity of each block by the activity calculation means, and calculates the first quantization table according to this activity. A second quantization table is created by the means. Using this second quantization table, the transform coefficients obtained by orthogonal transformation by the orthogonal transform means are quantized. As a result, optimal quantization is performed by changing the quantization table according to the activity of each block, and efficient encoding becomes possible.

0010

[Embodiments] Hereinafter, embodiments of the present invention will be explained with reference to the drawings. FIG. 1 shows a block diagram of an encoding device according to a first embodiment of the present invention. 1 as a component as shown in Figure 1.
01 is a DCT unit that performs orthogonal transformation on input image data to obtain transformation coefficients; 102 is a DC
a buffer memory as a coefficient storage means for temporarily storing the transform coefficients obtained by orthogonal transformation in the T section 101; an activity calculation section 105 as an activity calculation means for calculating the activity in the block from the image data;
106 is a quantization factor control unit as a quantization factor control means that controls the quantization step based on the activity obtained by the activity calculation unit 105; 108 is a quantization table that is a first quantization table; 107 is a quantization factor A calculation unit 103 is a calculation unit that performs calculations between the quantization factor obtained by the control unit 106 and the quantization table 108, and a calculation unit 103 uses the quantization table, which is the second quantization table obtained by the calculation unit 107. A quantization section 104 serves as a quantization means for quantizing the transform coefficients temporarily stored in the buffer memory 102; 104 is a quantization section 103;
This is an encoding unit that encodes the transform coefficients obtained in . The operation of FIG. 1 will be explained in detail below.

[0011] Here, the activity calculation unit 105 performs the calculation shown in (Equation 1). (Math. 1) is 8
It shows the definition of a block of ×8 pixels. In (Equation 1), i and j indicate the horizontal and vertical positions in the image, respectively, and l and m indicate the horizontal and vertical positions of the image data within the block. Activity ACT
differs for each block depending on the image. For example, if the block is on the edge of the image, the ACT will be large, and if the block is on the flat part, the ACT will be small.

0012

[Math 1]

[0013] where x: pixel in block DC: average value of block (i, j).

Next, the quantization factor control section 1 in FIG.
The circuit operation of 06 will be explained in detail using FIG. FIG. 2 shows a case where four types of quantization factors can be set. As shown in FIG. 2, the components include a threshold setting section 401 that sets a threshold, and a comparison section 402 that compares the input activity α with the threshold set by the threshold setting section 401. , 404 to 407 are quantization factor setting units that set quantization factors, and 403 is a selector unit that selects a quantization factor according to the result determined by the comparison unit 402. The input activity α is compared by a comparator 402 with a threshold value set in a threshold setting unit 401 . The selector unit 403 selects the quantization factor setting unit 40 according to the result of the comparison unit 402.
Select one from 4 to 407 and output the quantization factor.

Further, the circuit operation of the arithmetic unit 107 in FIG. 1 will be explained in detail using a quantization table.

[0016]

[Table 1]

[0017]

[Table 2]

[0018] The values shown in (Table 1) and (Table 2) are quantization tables corresponding to blocks of 8 x 8 pixels, and the image data at the same position in the block is divided by each element in the table. do. (Table 1) is the quantization table 10
(Table 2) is the calculation result when the output from the quantization factor control unit 106 is 1/2. In FIG. 1, the transform coefficients in the buffer memory 102 are quantized using the quantization table shown in Table 2.

Note that the activity calculation unit 105 can also divide each block into arbitrary subblocks. For example, each block can be divided into four subblocks as shown in Figure 3, the activity for each subblock can be calculated using (Equation 2), and the activity of the block can be calculated more accurately by adding them. It is possible to set the quantization factor suitable for the block, and it is possible to compress efficiently.

[0020]

[Math 2]

In (Equation 2), i and j indicate the horizontal and vertical positions of the block in the image, and l and m indicate the horizontal and vertical positions of the image data within the block. Also, the first term in (Equation 2) is subblock 0 in FIG. 3, the second term is subblock 1 in FIG. 3, the third term is subblock 2 in FIG. 3, and the fourth term is subblock 3 in FIG. The activity values of each are shown.

As shown above, the coding efficiency of performing quantization using the quantization table obtained according to the present invention is improved. This will be explained below using figures. FIG. 4 shows a method of encoding transform coefficients, and FIG. 5 shows coefficients after quantization and their encoding in the present invention and a conventional example. In Figure 4, when encoding transform coefficients, a zigzag scan is performed as shown in the figure, and the combination of continuous length of zero coefficients (zero run) and non-zero is encoded by converting it into a one-dimensional transform coefficient sequence. . In FIG. 5, white circles indicate zero conversion coefficients, and black circles indicate non-zero conversion coefficients.
Conventionally, as shown in FIG. 5(a), the code consists of three zero-run codes up to transform coefficient a, up to transform coefficient b, and up to the end of the block. On the other hand, in the present invention, FIG. 5(b)
As shown in the figure, the code is composed of two zero-run codes, one up to the transform coefficient a and the other up to the end of the block. As described above, according to the present invention, encoding efficiency can be improved.

A block diagram of an encoding device in a second embodiment of the present invention is shown. As shown in the figure, 201 is a DCT unit that performs orthogonal transformation on input image data to obtain transformation coefficients, and 202 is a buffer memory that temporarily stores the transformation coefficients obtained by orthogonal transformation in the DCT unit 201. , 205 is an activity calculation unit that calculates the activity within a block from the transform coefficient obtained by orthogonal transformation in the DCT unit 201, and 206 is an activity calculation unit 205.
208 is a quantization table, which controls the quantization step based on the activity obtained in
207 is a calculation unit that performs calculations between the quantization factor obtained by the quantization factor control unit 206 and the quantization table 208; 203 is a calculation unit that uses the quantization table obtained by the calculation unit 207 to temporarily store data in the buffer memory 202; A quantization unit 204 is a quantization unit 203 that quantizes the transform coefficients.
This is an encoding unit that encodes the transform coefficients obtained in . The related operations of the components shown in FIG. 6 will be described in detail below.

[0024] In the activity calculation unit 205,
An area is arbitrarily set for the transform coefficients temporarily stored in the buffer memory 202, and the mean square value of the transform coefficients included in that area is calculated. FIG. 7 shows an example of area setting. FIG. 7 shows transform coefficients within a block, and information is concentrated in the direction of the upper left part. In this case, the mean square value of the conversion coefficients in the shaded area in FIG. 7 is calculated, and that value is used as the activity value.

In the quantization factor control section 206, as shown in FIG.
Performs the same processing as the quantization factor control unit 106 of
A quantization factor is set according to the activity value calculated by the activity calculation unit 205 and output.

The calculation unit 207 performs the same processing as the calculation unit 107 in FIG. 1, generates a quantization table according to the output from the quantization factor control unit 206,
Quantize with 3.

Although the first and second embodiments have been described using discrete cosine transform as the orthogonal transform, it goes without saying that other orthogonal transforms such as Hadamard transform can have similar effects.

[0028]

As is clear from the above description, the encoding apparatus of the present invention has the following effects by providing orthogonal transformation means and arithmetic means for obtaining a quantization table according to the activity of each block.

(1) Since quantization is performed by changing the quantization table according to the activity of each block, efficient encoding is possible.

(2) Since processing is performed in blocks, activity calculations and code processing can be performed in parallel, resulting in high-speed processing.

[Brief explanation of the drawing]

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

[Figure 2] Block diagram of the quantization factor control unit of the same embodiment

[Fig. 3] Conceptual diagram of division of each block of an image signal into subblocks in the same embodiment.

[Fig. 4] Transform coefficient array pattern diagram for explaining encoding in the same embodiment.

[Fig. 5] Transform coefficient array pattern diagram showing an encoding example in the same embodiment.

FIG. 6 is a block diagram of an encoding device according to a second embodiment of the present invention.

[Fig. 7] Conversion coefficient array pattern diagram for explaining the operation of the same embodiment.

[Fig. 8] Block diagram of a conventional encoding device

[Fig. 9] A conversion coefficient array pattern diagram for explaining the operation of the conventional example.

[Explanation of symbols]

101 DCT section 102 Buffer memory 103 Quantization section 105 Activity calculation section 106 Quantization factor control section 107 Arithmetic unit 108 Quantization table

Claims (2)

[Claims] 1. Orthogonal transformation means for dividing image data into blocks of N×N pixels (N is a natural number) and performing orthogonal transformation on the image data, and coefficients for temporarily storing the results obtained by the orthogonal transformation means. a storage means, an activity calculation means for calculating an activity representing the fineness of an image in the block from the image data, and a table for storing a first quantization table for controlling the quantization step width of each coefficient in the block. a storage means, an arithmetic means for multiplying the first quantization table by a quantization factor to obtain a second quantization table, a quantization factor control means for obtaining the quantization factor by an activity; quantization means for quantizing the coefficients in the coefficient storage means using a quantization table, the activity calculation means calculates an activity for each block, and the quantization factor control means uses the activity to quantize the coefficients. A quantization factor is obtained, a second quantization table is obtained by the calculation means using the quantization factor and the first quantization table, and a second quantization table is obtained by the quantization means using the second quantization table. An encoding device that performs quantization. 2. Orthogonal transformation means for dividing image data into blocks of N×N pixels (N is a natural number) and performing orthogonal transformation on the image data, and coefficients for temporarily storing the results obtained by the orthogonal transformation means. a storage means, an activity calculation means for calculating the activity in the block from the result obtained by the orthogonal transformation means, and a table for storing a first quantization table for controlling the quantization step width of each coefficient in the block. a storage means, an arithmetic means for multiplying the first quantization table by a quantization factor to obtain a second quantization table, a quantization factor control means for obtaining the quantization factor by an activity; quantization means for quantizing the coefficients in the coefficient storage means using a quantization table, the activity calculation means calculates an activity for each block, and the quantization factor control means calculates an activity using the activity. A quantization factor is obtained, a second quantization table is obtained by the calculation means using the quantization factor and the first quantization table, and a second quantization table is obtained by the quantization means using the second quantization table. An encoding device configured to quantize the blocks.