JPH03243066A - Image coding system - Google Patents

Abstract

PURPOSE:To make the distortion of respective blocks inconspicuous and to suppress the deterioration of an image by dividing inputted image data into non-square blocks and quantizing cosine-converted data. CONSTITUTION:An input image is stored in an image memory 1 and divided into non-squrae blocks by a dividing circuit 2. Each divided block image is consine-converted by a conversion circuit 3, the cosine coefficient is quantized by a quantizing circuit 4 to reduce data volume and the quantized data are stored in a storage device 5. Since the divided blocks are formed as non-squares, the boundary of respective blocks is inclined and the distortion of each block is made inconspicuous. Since the divided blocks are quantized, the generation of block distortion can be suppressed on the receiving side in the case of executing the quantization of a low bit rate, suppressing the deterioration of an image.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image encoding method applied to an image compression/decompression device in an electronic filing system or the like, facsimile transmission, a video telephone, an electronic publishing system, and the like.

[Conventional technology]

One of the image encoding methods is transform encoding.

In transform encoding, a screen is divided into blocks of an appropriate number of pixels, and a numerical string consisting of sample values is orthogonally transformed for each block. As a result, power is concentrated on specific components due to the statistical properties of the image. Therefore, a large amount of code is assigned to components with high power, and components with low power are quantized with a small number of bits.

Discrete cosine transform (
DCT) is well known.

Conventionally, in coding that performs these block divisions, the blocks are divided into square blocks such as 8×8 as shown in FIG. However, these blocks had a problem in that the boundaries between the blocks were conspicuous. To solve this problem, a method has already been proposed to make the boundaries of blocks uneven (
JP-A-1-225293).

[Problem to be solved by the invention]

However, in the above-mentioned prior art, there is a problem that the boundary in the horizontal direction is a straight line and the boundary is conspicuous.

In addition, since the transform encoding shown earlier in the above-mentioned process divides the data into plots and performs encoding, when encoding at a low bit rate, the block distortion that occurs on the receiving side is the biggest cause of image deterioration. It has become.

An object of the present invention is to provide an image encoding method in which block distortion is not noticeable and image deterioration can be suppressed.

[Means to solve the problem]

The above objective is achieved by dividing manually input image data into non-square blocks in an image encoding method that divides image data input from an image input device into appropriate blocks and encodes each block. Ru.

Furthermore, this can also be achieved by making the non-square blocks described in - to have mutually different phases.

[Effect]

According to the first method, the blocks to be divided are made into non-square shapes, so that the boundaries of the blocks are diagonal, so that the distortion is not visually noticeable.

According to the second method, in addition to the above, by alternating the inclination directions of the diagonal blocks, encoding distortion is further made less noticeable visually.

〔Example〕

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

FIG. 1 is a block circuit diagram in which the image encoding method according to the present invention is implemented, in which 1 is an image memory that stores an input image, 2 is a dividing circuit that divides the image into blocks, and 3 is a divided block. A conversion circuit that cosine transforms an image, 4 a quantization circuit that quantizes the converted cosine coefficients to reduce the amount of data, and 5 a storage device that stores quantized image data.

Here, the cosine transform encoding method applied to the present invention will be described. In the cosine transform encoding method, first the image is N
Divide into XN blocks, and sequentially create two-dimensional D for each block.
Information is compressed by performing CT transformation and quantizing the transformed coefficients.

(1) The input image signal is converted into blocks of NXN pixels,
Orthogonal transformation is performed by two-dimensional DCT transformation.

Two-dimensional DCT of image signal X; J (i, j = o to N-1)
The conversion is defined by the following equation.

2・N u, v=o, t, -----, N-1c (
w) −1; w=o, 1. 2. -N(2) Linear quantization The coefficients obtained by orthogonal transformation are each quantized. At this time, bit allocation for each coefficient is determined based on the variance value of each component. The variance value of each coefficient is K
K where K is the total number of blocks, and the number of bits allocated to each coefficient is given by: At this time, M indicates the number of bits allocated for one block. In this way, by quantizing coefficients with large variance with a large number of bits and quantizing small coefficients with a small number of coefficients, an image with less visual deterioration when reproduced can be obtained even if the overall number of bits is reduced. To restore encoded data, it is sufficient to perform the reverse process of encoding.

That is, coefficients are obtained by inverse quantization, and two-dimensional DCT inverse transformation is performed to reproduce the image.

A flowchart of the above encoding procedure is shown in FIG.

First, NXN data is sequentially read (101), two-dimensional DCT transformation is performed (102), the variance value of each coefficient is determined (103), and the bit allocation of each coefficient is determined (10
4). Next, read the NXN data sequentially again (105)
, two-dimensional DCT transformation is performed (106), and the coefficients are quantized based on the bit allocation obtained in the coefficient bit allocation (104) (107).

FIG. 3 is an explanatory diagram showing image division using the encoding method according to the first embodiment, in which image data input to the image memory 1 is divided into non-square blocks as shown in this diagram.

Then, the divided blocks are treated as one image and cosine-transformed by the transformation circuit 3, and the coefficients thereof are quantized by the quantization circuit 4 and stored in the storage device 5.

FIG. 4 is an explanatory diagram showing image division using the encoding method according to the second embodiment. The division of image data into non-square blocks is the same as in the first embodiment, but this embodiment In the example, as shown in FIG. 4, the inclination directions of the blocks are further alternated. The subsequent processing is the same as in the first embodiment.

〔Effect of the invention〕

As explained above, according to the invention described in claim 1, the blocks to be divided for encoding are made non-square, so that the boundaries of the blocks are diagonal, so that encoding distortion is not visually noticeable. can do.

Further, according to the second aspect of the invention, in addition to the above, by alternating the inclination directions of the blocks, it is possible to make encoding distortion even more visually inconspicuous.

[Brief explanation of drawings]

FIG. 1 is a block circuit diagram in which the image encoding method according to the present invention is implemented, FIG. 2 is a flowchart showing the encoding procedure, and FIG. 3 is a diagram showing image division using the encoding method according to the first embodiment. FIG. 4 is an explanatory diagram showing image division by the encoding method according to the second embodiment, and FIG. 5 is an explanatory diagram showing image division by the encoding method according to the conventional example. ■...Image memory, 2...Dividing circuit, 3...Conversion circuit, 4...Quantization circuit, 5...Storage device. Figure 2 Figure 3 Figure

Claims (2)

[Claims] (1) An image encoding method that divides image data input from an image input device into appropriate blocks and encodes each block, characterized by dividing the input image data into non-square blocks. Image encoding method. (2) The image encoding method according to claim 1, wherein the non-square blocks have mutually different phases.