KR20030035517A - Improving fractal image compression/recovery method and apparatus thereof - Google Patents

Abstract

PURPOSE: An improved method and apparatus for compressing and/or decompressing fractal images are provided to improve compression rate without deteriorating picture quality of a decompressed image. CONSTITUTION: A fractal image is encoded to generate a compression parameter. Prediction is performed for the generated compression parameter on the basis of parameter values of previous blocks adjacent to the current block, and a difference between the predicted value and parameter value of the current block is scalar-quantized. The compression parameter includes a scale parameter and an offset parameter. The scalar-quantization(340) is carried out in such a manner that the scale parameter is scalar-quantized, prediction is performed for the offset parameter based on parameter values of neighboring previous blocks, and a difference between the predicted value and parameter value of the current block is scalar-quantized.

Description

Improved fractal image compression / recovery method and apparatus

The present invention relates to a fractal image compression and / or decompression system, and more particularly, to an improved fractal image compression and / or decompression method and apparatus for applying quantization in a pridiction scheme.

In general, international standardization of video compression technology for encoding video at a very low bit rate for video telephony or video conferencing systems such as Moving Pictures Experts Group (MPEG-4) has been conducted. MPEG-4 uses low-speed transmission channels of less than 64 kbps to transmit video at very low bit rates, which is essential for high compression. Among these high compression algorithms, fractal compression methods that can reconstruct images independently of resolution are being studied.

1 is a block diagram of a conventional fractal image compression apparatus.

Referring to FIG. 1, the fractal encoder 110 performs known fractal image compression to generate compression parameters, that is, scales S, offsets, and positions Pos.

The memory 130 stores scale (S) offset parameters and position (Pos) parameters, which are compression parameters generated by the fractal encoder 110.

The vector quantizer 140 searches for a representative code existing at a position closest to the scale (S) parameter and the offset (Off) parameter in the table, and outputs an index of the determined representative code.

The controller 120 performs operations on the fractal encoder 110, the memory 130, and the vector quantizer 140 and controls transmission of compressed data.

2 is a block diagram of a conventional fractal image restoration apparatus.

Referring to FIG. 2, the vector dequantization unit 210 generates a corresponding representative code value, that is, a scale (S) parameter and an offset (Off) parameter, from a received index.

The memory 230 stores other received parameters Pos and Ect and a scale S parameter and an offset parameter generated by the vector dequantization unit 210.

The fractal decoder 240 reads the compression parameters S, Off, Pos, Etc .. from the memory 230 and restores the original image.

The controller 220 performs various controls necessary for the vector quantizer 210, the memory 230, and the fractal decoder 240 to operate.

In the fractal compression / restoration apparatus shown in FIGS. 1 and 2, the compression parameters S and Off are quantized by a vector quantization unit. However, in the fractal compression method, it is not preferable to use vector quantization because the compression parameter (S) has a great influence on whether the image is to be restored and the image quality of the restored image. In addition, vector quantization requires an additional memory compared to scalar quantization, and may have a deterioration in image reconstruction quality. On the other hand, even when scalar quantization of compression parameters S and Off (not shown) in a fractal compression / recovery device, there is a fixed number of bits without reflecting the correlation between the compression parameters and the characteristics of each parameter. There is a disadvantage in that the compression rate is reduced by assigning.

The present invention provides a method and apparatus for compressing and / or reconstructing a fractal image which combines scalar quantization and quantization by linear prediction to generate a compression ratio without degrading the quality of the reconstructed image. will be.

In order to solve the above technical problem, the fractal image compression method according to the present invention

Generating compression parameters by encoding the fractal image;

The compression parameter generated in the above process includes performing prediction based on parameter values of previous blocks neighboring the current block and scalar quantizing a difference between the prediction value and the parameter value of the current block.

In order to solve the above other technical problem, the fractal image restoration method according to the present invention

Predicting a compression parameter of the current block based on dequantized compression parameters of the current block and neighboring previous blocks;

And adding the prediction error value of the received compression parameter and the compression parameter predicted in the above process to generate an inverse quantization value of the compression parameter for the current block.

In order to solve the above other technical problem, the fractal image compression and / or decompression device according to the present invention

Compression means for performing prediction on the compression parameters generated by encoding the fractal image based on parameter values of previous blocks neighboring the current block and quantizing a difference between the prediction value and the parameter value of the current block;

Predict the compression parameter of the current block based on the dequantized compression parameters of the current block and neighboring previous blocks, and add the predicted error value of the compression parameter received from the compression means and the predicted compression parameter to the current block. And reconstruction means for generating an inverse quantized value of the compression parameter for.

1 is a block diagram of a conventional fractal image compression apparatus.

2 is a block diagram of a conventional fractal image restoration apparatus.

3 is a block diagram of a fractal image compression apparatus according to the present invention.

4 is a detailed view of an offset quantization unit of FIG. 3.

5 is a block diagram of a fractal image reconstruction device according to the present invention.

6 is a detailed view of an offset inverse quantization unit of FIG. 5.

7 is a diagram illustrating a positional relationship between a current block and a neighboring previous block in FIGS. 4 and 6.

Hereinafter, exemplary embodiments of the present invention will be described with reference to the accompanying drawings.

3 is a block diagram of a fractal image compression apparatus according to the present invention.

Referring to FIG. 3, the fractal encoder 310 performs fractal image compression on blocks divided into a predetermined size, thereby compressing parameters, that is, a scale parameter, an offset parameter, a position parameter, and the like. Create other parameters (Ect).

The memory 330 stores scale (S), offset (Off) parameters, position (Pos) parameters, and other parameters (Ect), which are compression parameters generated by the fractal encoder 310.

The scalar quantization unit 340 performs a scalar quantization using the number of bits given to the scale parameter S (i) of the current block read out to the memory 330 by the control unit 320 to quantize the scale parameter S (i ) ).

The offset quantization unit 350 calculates the S (i, k) and Off (i, k) parameters previously obtained with respect to the offset parameter (Off (i)) of the current block read by the controller 320 into the memory 330. Linear prediction is performed from the set O (i), and the value E (i) obtained by scalar quantization of the difference between the offset parameter Off (i) and the prediction value of the current block is transmitted.

The controller 320 performs various controls necessary for the fractal encoder 310, the memory 330, the scalar quantizer 340, and the offset quantizer 350 to operate, and controls the scale (S) and offset (Off) parameters. In addition, the compression parameters Pos and Etc are read from the memory 330 and transmitted.

4 is a detailed view of the offset quantization unit 350 of FIG. 3.

Referring to FIG. 4, first, the parameters represented by the set of input signals O (i) defined as Equation 1 are read from the memory 330 to quantize the offset parameters of the current block i.

O (i) = {Off (i), S (i), Off (i, 1), Off (i, 2), ...., Off (i, k), S (i, 1), S (i, 2), ...., S (i, k)}

Here, Off (i, k) and S (i, k) mean scale (S) and offset (Off) parameters for (k) th blocks adjacent to the (i) th block. An embodiment of the positional relationship between these blocks is shown in FIG.

The first scalar quantizer 410 and the second scalar quantizer 420 are S (i) and the previous block around the current block (i) obtained for the current block (i) of the set of signals (O (i)). The scalar quantization is performed on S (i, k) obtained for these fields to generate S (i) and S (i, k) , respectively.

The linear prediction unit 430 uses Equation 2 using Off (i, k), S (i) and S (i, k) obtained from previous blocks (i, k) around the current block i to be compressed. We predict the offset parameter ( Off (i) ) as

Off (i) = f {S (i), S (i, k), Off (i, k)}

Where k is an arbitrary coefficient value.

The third scalar quantizer 460 scalar quantizes the difference E (i) between the offset parameter Off (i) and the predicted offset parameter Off (i ) of the current block, and its value E (i). Obtain and send).

The offset parameter Off (i), which is the sum of the error quantization value E (i) and the predicted offset parameter Off (i), is stored in the memory 330 for use in the quantization of another block.

The same algorithm as the block shown in FIG. 4 is applied to all blocks.

5 is a block diagram of a fractal image reconstruction device according to the present invention.

Referring to FIG. 5, the scale dequantization unit 515 dequantizes the received quantization scale parameter S (i) .

The memory 530 is an inverse quantization scale parameter S of the current block received by the controller 520, other compression parameters Pos, Ect, and an offset of the current block generated by the offset dequantization unit 510. Save the parameter (Off (i)).

The offset dequantization unit 510 offsets the current block from the received error quantization value E (i) and the S (i, k) and Off (i, k) parameters previously dequantized and stored in the memory 530. The parameter Off (i) is generated, and the offset parameter Off (i) is stored in a designated place of the memory 530.

The fractal decoder 540 reads the compression parameters S , Off, Pos, and Etc from the memory 530 to restore the entire image.

The controller 520 performs various controls necessary for the offset dequantization unit 510, the memory 530, and the fractal decoder 540 to operate.

6 is a detailed view of the offset inverse quantization unit 510 of FIG. 5.

First, parameters (Off (i, 1), Off (i, 2) ... Off (i) of adjacent blocks having the same positional relationship as the current block i from the memory 520 are controlled by the controller 520. , k), S (i, 1) , S (i, 2) , ....., S (i, k) ) are read.

The linear prediction unit 620 calculates the offset parameter value Off (i) of the predicted current block by using the read S (i, k) and Off (i, k). The offset parameter value Off (i) of the predicted current block and the received quantization value E (i) are added to generate an inverse quantization value of the offset parameter for the current block i.

The same algorithm as the block shown in FIG. 6 is applied to all blocks, and the compression parameters are stored in the memory 530.

7 is a diagram illustrating a positional relationship between a current block (i) and a neighboring previous block (i, k) when k = 4 applied to FIGS. 4 and 6.

Referring to FIG. 7, when k = 4, the current block (i) and its neighboring four previous blocks ((i, 1), (i, 2), (i, 3), (i, 4)) In addition, the positional relationship of neighboring previous blocks may vary according to any coefficient k value.

As described above, according to the present invention, a compression ratio is performed by performing prediction using offset (S) and offset (Off) parameters obtained from the neighboring block of the block to which the parameter to be quantized is offset. In addition, by applying scalar quantization to the scale S parameter, it is possible to reduce the deterioration of image quality caused by the quantization of the scale S parameter.

Claims (12)

In the fractal image compression method, Generating compression parameters by encoding the fractal image; And performing a prediction on the compressed parameters generated in the above process based on the parameter values of previous blocks neighboring the current block and scalar quantizing a difference between the predicted value and the parameter value of the current block. The method of claim 1, wherein the compression parameter generation process And encoding the fractal image to generate scale and offset parameters. The method of claim 1, wherein the quantization process Scalar quantization of the scale parameter generated in the process, prediction based on parameter values of neighboring previous blocks with respect to the offset parameter, and scalar quantization of a difference between the predicted value and the parameter value of the current block. Fractal image compression method. A fractal compression decompression method for receiving quantized compression parameters and performing fractal image reconstruction, Predicting a compression parameter of the current block based on dequantized compression parameters of the current block and neighboring previous blocks; And generating an inverse quantization value of the compression parameter for the current block by adding the prediction error value of the received compression parameter and the compression parameter predicted in the process. The method of claim 4, wherein the compression parameter prediction process is performed. And predicting offset parameters of the current block based on dequantized scale and offset parameters of the current block and neighboring previous blocks. The method of claim 4, wherein the inverse quantization generation process is performed. And generating an inverse quantization value of the offset parameter for the current block by adding the prediction error value of the received offset parameter and the offset parameter predicted in the process. 5. The method of claim 4, wherein the inverse quantization generation process further comprises inverse quantization of a received quantized scale parameter. In the fractal image compression and / or decompression method, Encoding the fractal image is performed to generate scale and offset parameters, scalar quantizes the generated scale parameters, performs prediction based on parameter values of neighboring previous blocks, and predicts the offset parameters. Scalar quantization of the error between the current block and the current block; In the process, the quantized offset parameter and the scale parameter are received, and the offset parameter is predicted based on the dequantized scale parameter and the offset parameter with respect to the previous block and the previous block, and the quantized error value and Generating an inverse quantization value of an offset parameter for a current block by adding the predicted offset parameter. A fractal image compression and / or decompression device, Compression means for performing prediction on the compression parameters generated by encoding the fractal image based on parameter values of previous blocks neighboring the current block and quantizing a difference between the prediction value and the parameter value of the current block; Predict the compression parameter of the current block based on the dequantized compression parameters of the current block and neighboring previous blocks, and add the predicted error value of the compression parameter received from the compression means and the predicted compression parameter to the current block. And decompression means for generating inverse quantization values of the compression parameters for the fractal compression and / or decompression device. The method of claim 9, wherein the compression means A fractal encoder for generating a scale and offset parameter by performing fractal image compression on blocks divided into a predetermined size; A scalar quantizer configured to scalar quantize the scale parameter generated by the fractal encoder; Fractal compression and / or including an offset quantization unit for obtaining a prediction value based on the scale and offset parameters of the previous block with respect to the offset parameter generated by the fractal encoder, and scalar quantization of the difference between the offset parameter and the prediction value of the current block Or restore device. The method of claim 10, wherein the offset quantization unit A first scalar quantization unit configured to scalar quantize the scale parameter obtained for the current block to be compressed and the scale parameter obtained for previous blocks around the current block; A prediction unit for predicting the offset parameter based on the offset parameter obtained in the previous blocks around the current block to be compressed and the scale parameter obtained in the first scalar quantization; And a second scalar quantizer for scalar quantizing a difference between an offset parameter of the current block and a predicted offset parameter. The method of claim 9, wherein the restoring means A scale inverse quantizer for inversely quantizing the received quantized scale parameter; An offset inverse quantizer for generating an offset parameter of a current block from a received error quantization value and a scale and offset parameter previously dequantized and stored; And a fractal decoder which decodes the offset parameter generated by the offset inverse quantizer together with other compression parameters to restore the image signal.