KR19980023632A - Video signal compression coding device - Google Patents

Abstract

The present invention relates to a video signal compression encoding apparatus, comprising: block means for blocking an input video signal into an orthogonal transform block, a small block, and a large block; Orthogonal transform means for performing orthogonal transform of the signal output from the blocking means in orthogonal transform block units; Quantization step size selection means for determining the quantization step size through a multi-stage encoding process by inputting the orthogonal transform coefficients output from the orthogonal transform means; A buffer for delaying the orthogonal transformed coefficients output from the orthogonal transformation means while the quantization step size selection means selects the quantization step size; Quantization means for performing quantization on coefficients output from the buffer at the quantization step size selected by the quantization step size selection means; Continuous field encoding means for performing continuous field encoding on the quantized signal from the quantization means; Variable length encoding means for performing variable length encoding on the signal output from the continuous field encoding means; And a data sorter for rearranging the data in order to record the signal output from the variable length encoding means in a predetermined format. Accordingly, in order to increase the efficiency of the fast search, compression and coding of the input signal are performed in independent units, and the quantization can be performed more efficiently to adjust the code amount of the input signal to a certain size, thereby improving the image quality. .

Description

Video signal compression coding device

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video signal compression encoding apparatus, and more particularly, to a method for generating a quantization step size used in a signal compression step having a great effect in terms of image quality, and a device suitable therefor.

The advent of the Digital Video Cassette Recorder (hereinafter referred to as DVCR) has been made possible by the increasing recording density of video tapes, the improvement of digital compression techniques and the development of digital circuit technology. Since DVCR has many advantages such as maintaining the original image quality without deterioration even after repeated playback, conversion from an analog video tape recorder to a digital video tape recorder is inevitable. As the international standard for DVCR of standard video image is decided, much demand of DVCR is expected in the future and its application is also diversified.

The generation technique of the quantization step size used in the compression stage, which greatly affects the performance of the DVCR system, especially in terms of the image quality, is the key to the image quality of the entire system because most specifications are fixed in the compression method of the video signal. Part.

Conventionally, since a digital video signal must be recorded in a predetermined number of video tracks every frame when recording on a videotape, many problems occur in compressing and recording a video signal using a variable length coding method in which the output is not constant. That is, the compressed and output video signal may be smaller than the fixed target length in some cases and may be exceeded in some cases. In this case, since the degradation and damage of the image quality to be reproduced are caused, a code length output method approaching a fixed target length by using a given area as efficiently as possible is essential in the video signal compression part of the DVCR.

In addition, among the additional functions of the DVCR, high-speed search is essential and the implementation of this function is one of the most difficult problems. In the case of an analog video tape recorder, there is no need for a high speed search. However, in the case of DVCR, recording is performed using the variable length coding method, so that even if a part is read, all signals cannot be reproduced. Currently, in order to implement such a fast search function, it is necessary to fix a code code that is compressed and output to a unit capable of independent decoding of as few units as possible.

To this end, variable-length coding is performed in units of one large block composed of a certain number of small blocks, and the output code codes are forcibly fixed in accordance with a predetermined format for recording in a constant video tape area. The performance of fast searching can be improved by reducing the independent coding units from the entire frame to the sub-units. However, although the performance of the large block unit may be good for high-speed search, there is a relative loss in image quality compared to the frame unit in terms of signal compression.

In a method for recording a code encoded in one independent large block unit on a certain area of a video tape, the most important part of the performance is a quantization process. The compression process is largely composed of a variable length coding process and a quantization process. In the variable length coding process, a signal that has undergone continuous length coding (RLC) is encoded by a variable length code table, so that it is changed according to an input signal. It is a compression-free compression process that only encodes. In the previous stage of the process, an accurate and efficient quantization process used to reduce the amount of data is important and essential for recording in a fixed area of a certain size according to the amount of information of the input signal.

The present invention is to provide a video data compression technique for a digital video signal to be reproduced for a sufficient time on a video tape of a suitable size, in particular, a method of compression coding the input signal in an independent unit in order to increase the efficiency of fast search To provide a more efficient video signal compression coding apparatus for matching the code amount of the input signal to a constant size.

1 is a block diagram of a video signal compression encoder for recording a compression coded signal in a fixed size region according to the present invention.

FIG. 2 is a diagram showing the configuration of the blocking unit 2 of FIG.

3 is a detailed block diagram of a quantization step size selection unit as an embodiment for selecting a quantization step size according to the present invention.

4 is a detailed block diagram of a quantization step size selection unit as another embodiment for selecting a quantization step size according to the present invention.

5 is a block diagram of a quantizer according to the present invention.

6 is a flowchart of one embodiment showing selection of a quantization step size in accordance with the present invention.

7 is a flowchart of another embodiment showing selection of a quantization step size in accordance with the present invention.

8 is a flowchart illustrating a method of selecting a quantization step size according to the present invention.

In accordance with another aspect of the present invention, there is provided a video signal compression encoding apparatus comprising: block means for blocking an input video signal into an orthogonal transform block, a small block, and a large block; Orthogonal transform means for performing orthogonal transform of the signal output from the blocking means in orthogonal transform block units; Quantization step size selection means for determining the quantization step size through a multi-stage encoding process by inputting the orthogonal transform coefficients output from the orthogonal transform means; A buffer for delaying the orthogonal transformed coefficients output from the orthogonal transformation means while the quantization step size selection means selects the quantization step size; Quantization means for performing quantization on coefficients output from the buffer at the quantization step size selected by the quantization step size selection means; Continuous field encoding means for performing continuous field encoding on the quantized signal from the quantization means; Variable length encoding means for performing variable length encoding on the signal output from the continuous field encoding means; And a data sorter for rearranging the data in order to record the signal output from the variable length encoding means in a predetermined format.

Blocking means for the equalization of the input information is characterized in that it is possible to process in a small block unit after the processing of a small block unit of several small blocks or processing of a small block unit after the processing of a small block unit consisting of several orthogonal transform blocks. It is done.

The quantization means is controlled by the quantization step size and characterized in that the input signal is bit-shifted or divided by the step size multiplied by the quantization table divided by the input signal.

The quantization step size selecting means performs quantization means, continuous field coding means, variable length coding means, and accumulation means in parallel using several quantization step sizes in addition to the reference quantization step size to output a result closest to the target information amount. It is characterized by selecting a step size.

The normalizer of the quantization step size selector performs a quantization process on a quadrature transformed coefficient that is a given input signal using a reference quantization step size, and then calculates the number of symbols in orthogonal transform block units generated through a continuous field coding process. The number of new symbols available for each small block is allocated by using a normalization relationship between the sum of the small blocks and the large blocks and the number of symbols corresponding to the usable fixed lengths.

The normalizer of the quantization step size selector uses a reference quantization step size with respect to the quadrature transformed coefficients, which are given input signals, and uses code lengths of orthogonal transform blocks generated through quantization, continuous field coding, and variable length coding. It is characterized by allocating a new code code length that can be used for each small block by using the normalization relationship between the sum of the small block unit and the large block unit and the usable fixed length unit of the large block unit.

The determination of the quantization step size of the quantization step size selector is the actual quantization of the quantization step size which outputs the result with the least error among the values assigned to each small block and the output value by several quantization step sizes by a normalization process. It is characterized by using in a group.

Hereinafter, the present invention will be described in more detail with reference to the drawings.

1 is a block diagram of a video signal compression encoder for recording a compression coded signal in a fixed size region according to the present invention.

In Fig. 1, reference numeral 1 is an input signal, 2 is a block unit, 3 is an orthogonal transform unit, 4 is a quantization step size selection unit, 5 is a buffer, 6 is a quantization unit, and 7 is a continuous field. An encoding unit, 8 is a variable length encoding unit, 9 is a data sorter unit, and 10 is an output signal.

When the digital image signal 1 is input, an orthogonal transform block that is an execution unit of orthogonal transform coding, a block of several small blocks and a block of large blocks of several small blocks are performed. . The blocking unit 2 equally divides the information of the digital video signal input over one frame for the purpose of compression.

FIG. 2 is a diagram showing the configuration of the blocking unit 2 of FIG.

The minimum unit of blocking is the orthogonal transform block 11, and the small block 12 composed of several orthogonal transform block units passes through the large block unit 13 composed of several small blocks. The small-blocking unit 12 and the large-blocking unit 13 shown in FIG. 2 may be configured in the reverse order shown in the construction method, that is, in the order of the large-blocking unit 13 and the small-blocking unit 12. .

As described above, when an input of an independently decodable unit is input in a compression encoding process, an orthogonal transformation is performed first. The most representative example of the orthogonal transformation is the most commonly used discrete cosine transform (DCT). The coefficients with orthogonal transformation are performed through two paths. One is composed of a quantization step size selection unit 4 used in the quantization process, and the other is composed of a buffer 5 for delaying coefficients subjected to orthogonal transformation while the quantization step size is obtained.

In the compression-coding of the input signal at a constant compression ratio, the quantization means of the quantization unit 6 plays the most important role in outputting the compression-coded data with a substantially constant size for various input signals. In the case of continuous length coding of the continuous length coding unit 7 and variable length coding of the variable length coding unit 8 shown later, the amount of code is reduced to some extent without error, but the amount thereof is limited. There is no choice but to rely on adjustment.

Coefficients subjected to quantization by the selected quantization step size perform a continuous field coding process. Continuous field coding is a symbol consisting of a combination of nonzero values and counting the number of coefficients input in units of orthogonal transform blocks until a nonzero value is encountered. ) The variable length coding unit 7 assigns a code word having a small code length to a symbol having a high probability of occurrence for a symbol output from the continuous field coding unit 7, and applies a symbol word having a low probability of occurrence An encoding process is performed by allocating codewords having a long code length.

The data output from the variable length coding unit 8 is not constant in length, and is not constant in output period. Therefore, the data output from the variable length coding unit 8 must be aligned in order to record the compressed coded data in a fixed sized area according to a given format. The data sorter unit 9 serves to convert the code code output from the variable length coding unit 8 into a word of a constant length and to record it in a fixed fixed area.

A method of selecting a quantization step size for performing quantization in the compression encoding process as described above will now be described.

3 is a detailed block diagram of the quantization step size selection unit 4 as an embodiment for selecting the quantization step size according to the present invention.

In Fig. 3, reference numeral 13 is a multistage quantizer, 14 is a multistage continuous field encoder, 15 is a multistage accumulator, 16 is a normalizer, and 17 is a quantization step size detector. The signal input to the multilevel quantizer 13 is a signal output from the orthogonal transform coding unit 3 as shown in FIG. 1, and the signal output from the quantization step size detector 17 is a quantization unit as shown in FIG. 1. Is outputted as (6).

Since the code amount output for each large block unit composed of several small blocks must be recorded in a fixed area, the quantization step size is determined for each orthogonal transform block, the quantization step size is determined for each small block, and the entire large block. Quantization step size can be determined.

Given a quantization step size per small block; The multilevel quantizer 13 performs quantization of the input orthogonal transform blocks by a fixed number of quantization step sizes. In the quantization process, as shown in FIG. 5, the orthogonal transformed coefficients are bit-shifted according to the quantization step size to reduce the values, or the quantization table corresponding to the orthogonal transform block is used to determine the table by the table. Quantization is performed by multiplying the coefficients of. Next, parallel continuous field coding 14 is performed on coefficients of each orthogonal transform block unit output by each quantization step size. The multi-stage accumulator 15 accumulates and outputs the output symbols of continuous field coding in units of orthogonal transform blocks for each small block unit. The normalizer 16 allocates a new target symbol value for each small block unit by using a symbol value accumulated in a reference large block unit and a symbol value accumulated in a small block unit. The quantization step size detector 17 selects a quantization step size that results in the smallest difference between the target symbol value given in small block units and the symbol value in small block units output in parallel by several quantization step sizes. . This selected quantization step size is actually used in the quantization process.

As another example, FIG. 4 is an input value used for the normalizer as shown in FIG. 3, and the output of the continuous field encoder is calculated using a variable length encoding process to calculate the actual code length and used in the normalizer. do. That is, a target code length is allocated for each small block using a code length that is a more accurate value rather than a symbol value of a continuous length encoder.

The method of selecting the quantization step size will be described in detail with reference to FIG. 6.

First, orthogonal transformed coefficients are input to the quantizer. The amount of code output in units of large blocks must be output in accordance with this size due to the limitation of recording in a fixed area, and thus, this purpose is achieved by adjusting the quantization step size in units of small blocks. In this process, the most important thing is how to approach the quantization step size closest to a fixed size per small block. In the present invention, a symbol of continuous length coding is obtained by using several quantization step sizes simultaneously, or a code length of variable length coding is obtained, and a normalization stage using the same is assigned a target value for each small block, and this value and several quantizations previously performed. The actual quantizer uses a quantization step size that produces the closest result when comparing the output value by the step size.

In the reference quantizer 34, the intermediate value between the quantization step sizes used in the upper multistage quantizer 35 and the quantization step sizes used in the lower multistage quantizer 33 is determined. Used as a value, a new target value is set for each small block.

As described above, the output value of the quantizer for each quantization step size is a symbol is output through a continuous field coding process in units of orthogonal transform blocks, and these symbol values are the accumulators 39, 40, and 41 of small units. It may be input to or as a variable length encoder as shown in FIG. In the first case, the output of the continuous-length encoder is used by the normalizer, and in the latter case, the output of the variable-length encoding is used.

In Fig. 6, the lower and upper multi-stage accumulators accumulate the output symbol values of the continuous field coder for each quantization step size in small blocks and enter the quantization step size selector 44, which is based on the reference quantization step size. The output symbol values of the continuous field encoder are accumulated in small block units, and the large accumulator 42 accumulates several small block unit symbol values. These values are used in the normalizer 43 to reassign the target symbol value for each small block.

The principle of operation of the normalizer is as follows.

When the final output based on the reference quantization step size is summed in units of large blocks, the output should not exceed a certain size when the compressed coded data should be recorded in a certain area. In this case, it is necessary to reduce the code amount by widening the quantization step size as a whole. On the contrary, when the output value is smaller than the target value, there is a part that is not recorded. Therefore, the code amount is increased, that is, the quantization step size is reduced to reduce the code amount. Compression encoding must be performed in a lesser direction to prevent deterioration of image quality during decoding.

In the normalizer, a reference value for determining a quantization step size for maximizing use of a recordable area in order to prevent deterioration of image quality in decoding is thus determined. In other words, if the total output value of the large block unit by the reference quantization step size is larger than the target value, the respective code amount is reduced in small block units to bring it closer to the overall target value. Reduce the size to adjust the overall fixed size target. As described above, the selection of the quantization step size of each small block unit is most important in increasing or decreasing the code amount by adjusting the quantization step size of the small block unit.

Some of the small blocks in the large block have more outputs than others, and some of them are smaller. When the output value of the large block unit exceeds the target value, the quantization step size is selected in which the decreasing amount of the small block having a relatively large output value is also reduced compared to other small blocks. This is because a relatively large amount of information is less sensitive to human visual characteristics than it would otherwise, so reducing the output of a more complex small block rather than reducing the output of a small block in a simple area reduces the degradation of image quality during decoding. Can be.

Therefore, through the normalization process to which this principle is applied, an appropriate quantization step size can be selected for each small block to select a more efficient quantization step size that can reach a target value.

Examples of the reference value used in the normalization process include the use of the output of the continuous length encoder as shown in FIG. 6 and the use of the output of the variable length encoder as shown in FIG. 7. In the former case, since the variable length encoder is omitted, the amount is reduced in terms of hardware. However, since the data actually recorded is the output of the variable length encoder, some errors may occur. On the other hand, when the output value of the variable length encoder is used, a more accurate quantization step size can be obtained.

When the output value of the continuous length encoder is used, since the output value of the encoder is proportional to the output value of the actual variable length encoder, the variable length encoder can be omitted to some extent by using this relationship.

First, the normalizer shown in FIG. 6 performs a normalization process of allocating the number of target symbols corresponding to the available code lengths for each small block. Here, as shown in the normalization equation 1 used, normalization is performed by the number of symbols of the large blocks output in the quantization process based on the reference quantization step size and the number of symbols of each small block constituting the large block, respectively. Reassign the number of available target symbols for every small block of.

Here, D is the number of symbols corresponding to a fixed code length that can be used as a large block, but does not always maintain the same value as the fixed code length and the number of symbols, but uses a value obtained by taking an average value from various input video signals. .

On the other hand, in the normalizer shown in FIG. 7 is a process of allocating usable code lengths for each small block, which corresponds to Equation 2 below.

The A, B, C, D and A ', B', C ', D' is as follows.

A: the number of usable symbols allocated for each small block

B: Number of symbols of a large block by reference quantization step size

C: number of small blocks by reference quantization step size

D: the number of target symbols corresponding to the available fixed code length corresponding to the large block

A ': Usable sign length allocated for each small block

B ': code length of large block by reference quantization step size

C ': Code length of small block by reference quantization step size

D ': Usable fixed code length for large blocks

Information through the normalizer selects a quantization step size that results in the least error as shown in FIG. 8. That is, a comparison is made between the output values of the small block unit by each quantization step size and the value assigned to each small block by the normalization equation, thereby selecting the quantization step size which results in the lowest absolute value of the difference value. The quantization step size determined through this process is actually used as the quantization step size of the quantization process.

The structure of the quantization step size selector of FIGS. 6 and 7 is as follows.

In the case of FIG. 6, the code length information 63 is the number of symbols in small blocks based on respective quantization step sizes (reference quantization step sizes, upper quantization step sizes, and lower quantization step sizes). The target code length information 64 is an assigned code length information value for each small block unit that has been normalized by a reference quantization step size, and the code length information error estimation ( 65 is an operator for subtracting two input signals, and the quantization step size estimation 66 is an operator for selecting a quantization step size that results in the least of these error values.

The video signal compression encoding apparatus as described above compresses and encodes an input signal in a predetermined unit to increase the efficiency of the fast search, and can perform more efficient quantization for matching the code amount of the input signal to a constant size. Provides the effect of improving image quality.

Claims (7)

Blocking means for blocking the input video signal into an orthogonal transform block and a small block and a large block; Orthogonal transform means for performing orthogonal transform of the signal output from the blocking means in orthogonal transform block units; Quantization step size selection means for determining the quantization step size through a multi-stage encoding process by inputting the orthogonal transform coefficients output from the orthogonal transform means; A buffer for delaying the orthogonal transformed coefficients output from the orthogonal transformation means while the quantization step size selection means selects the quantization step size; Quantization means for performing quantization on coefficients output from the buffer at the quantization step size selected by the quantization step size selection means; Continuous field encoding means for performing continuous field encoding on the quantized signal from the quantization means; Variable length encoding means for performing variable length encoding on the signal output from the continuous field encoding means; And And a data aligner for rearranging data in order to record a signal output from the variable length coding unit in a predetermined format. According to claim 1, Blocking means for the equalization of the input information is a small block unit after the processing of a small block unit or a small block unit of a plurality of small blocks after processing of a small block unit consisting of several orthogonal transform blocks A video signal compression coding apparatus, characterized in that the processing is possible. The video signal compression encoding apparatus of claim 1, wherein the quantization means is controlled by a quantization step size and bit-shifts the input signal or divides the input signal by multiplying the quantization table by the step size. The method according to claim 1, wherein the quantization step size selecting means performs quantization means, continuous field encoding means, variable length encoding means, and accumulation means in parallel using several quantization step sizes in addition to the reference quantization step size to obtain the closest target information amount. And a quantization step size for outputting the result. 5. The orthogonal transform block of claim 4, wherein the normalizer of the quantization step size selector generates a quantization process using a reference quantization step size for orthogonal transformed coefficients that are a given input signal, and then generates an orthogonal transform block. Allocate the number of new symbols available for each small block using a normalization relationship between the sum of the number of symbols in the unit, the small block unit and the large block unit, and the number of symbols corresponding to the available fixed length in the large block unit. And a video signal compression encoding device. 5. The orthogonalizer of claim 4, wherein the normalizer of the quantization step size selector is orthogonal generated through a quantization process, a continuous field coding process, and a variable length coding process using reference quantization step sizes for orthogonal transformed coefficients which are given input signals. A new codecode length that can be used for each small block is allocated by using a normalization relationship between the sum of the code lengths of the transform block unit and the small block unit and the large block unit, and the usable fixed length unit of the large block unit. Image signal compression encoding apparatus. 7. The method of claim 5 or 6, wherein the determination of the quantization step size of the quantization step size selector results in the least error among the values assigned to each small block by the normalization process and the output values by several quantization step sizes. And a quantization step size for outputting the quantizer is used in an actual quantizer.