KR19990009352A - Compression coding apparatus and method - Google Patents

Abstract

The present invention relates to an apparatus and method for setting a quantization step size in intra coding. For the purposes of the present invention, a first process of generating a first quantization step size in units of slices according to the complexity of input image data, and DCT the input image data to classify the quantized image data DCT block into a plurality of regions according to image characteristics. And a third process of performing quantization at a second quantization step size obtained by subtracting the first quantization step size of the first process according to the number of DCTs in each region in the second process. According to the present invention, since there is no feedback loop, no buffer control is required, and since intra frame coding does not require an apparatus for inter frame, hardware size is reduced and intra frame coding is performed. There is an advantage that it is easy to edit.

Description

Compression coding apparatus and method

The present invention relates to a compression encoding apparatus and method, and more particularly, to an apparatus and method for setting a quantization step size in intra encoding of MPEG-2. In general, MPEG-2 has a data rate of 4-10 Mbits / sec, and aims at existing broadcast picture quality of NTSC, PAL, SECAM, or digital TV picture quality of CCIR601.

1 is a block diagram of a typical MPEG-2 compression encoding apparatus, a preprocessor 112 for filtering an input video signal, a motion estimator (ME) 114 for inter-image prediction, and a frame memory stored in units of frames 124, a motion compensator (MC) 126 that compensates for motion using the estimated motion vector, a DCT device 116 that transforms a DCT (Discrete Cosine Transform) for 8 × 8 blocks to obtain spatial interpolation, space Quantizer (Q) 118 to remove redundancy, inverse quantizer 120 to dequantize, IDCT 122 to inverse DCT, variable length coding (VLC) 128 to remove statistical redundancy, rate adjustment Forward analyzer 132, a buffer for transmitting data 130 is configured.

As shown in FIG. 1, compression coding of MPEG-2 employs intra frame and inter frame coding. Intra-frame coding is performed through the quantizer 118 and the VLC machine 128 after performing DCT in the DCT apparatus 116 without performing predictive encoding. Inter-frame coding performs predictive coding and includes P (Predictive) frame coding and Bidirectional (B) coding. The predictive encoding encodes a motion vector by detecting a motion between images through a motion estimator (ME) 114, a frame memory 124, and a motion compensator (MC) 126. Therefore, the P frame detects a motion vector based on the I frame, and the B frame detects a motion vector based on the P frame and the I frame. In addition, the amount of data generated in the buffer 130 is irregularly generated between frames and macroblocks within the frame according to the characteristics of the input image. In order to transmit the irregularly generated data through a transmission line having a constant transmission speed while maintaining high quality, a rate adjustment is required in the forward analyzer 132, and a feedback loop is formed to form a quantized scale vector, MQUANT (Modified quantizer). Adjust the value. This MQUANT is a quantization step size, which is one of the important factors for determining image quality. In MPEG-2, three steps are used to determine the quantization step size. In the first step, the overall complexity is obtained by measuring the global complexity, and the bit rate previously generated for each I, B, and P picture, the current target bit rate, and remaining in one group of picture (GOP) are obtained. Bit allocation is made for each picture number. In the second step, a reference MQUANT value in macroblock units is determined according to the fullness of the current buffer 130. In the third step, the complexity of the current macroblock is calculated, normalized, multiplied by a reference MQUANT value, and applied to quantization by calculating an MQUANT value applied to the actual macroblock. As described above, the encoding process of MPEG-2 performs inter-frame coding, and it is impossible to edit frame by frame due to a feedback loop due to generated bits of I, B, and P pictures, and hardware has a disadvantage.

The technical problem to be achieved by the present invention is an intra frame encoding apparatus that obtains a first MQUANT according to the complexity of a slice unit and determines a final quantization step size by using a second MQUANT obtained by dividing the first MQUANT and the DCT coefficient into a plurality of regions. To provide it.

Another technical problem to be solved by the present invention is an intra frame encoding method, in which a first MQUANT is obtained by the complexity of a slice unit, and a final quantization step size is determined using a second MQUANT obtained by dividing the first MQUANT and a DCT coefficient into a plurality of regions. Is to provide a way.

1 is a block diagram of a general compression encoding apparatus of MPEG-2.

2 is a block diagram showing a compression encoding apparatus according to the present invention.

FIG. 3 is a 4: 2: 0 picture format diagram for a conventional macroblock.

4 is a typical slice format diagram of one frame.

5A is a low DCT block coefficient format diagram.

5B is a horizontal edge DCT block coefficient format diagram.

5C is a vertical edge DCT block coefficient format diagram.

5D is a Diagonal Edge DCT block coefficient format diagram.

5E is a high DCT block coefficient format diagram.

6 is a flowchart of a compression encoding method according to the present invention.

In order to solve the above technical problem, the present invention provides an MPEG-2 compression encoding apparatus comprising: first quantization means for quantizing by a first quantization step size calculated with a complexity of a slice unit of input video data; Target bit calculating means for calculating a target bit in units of macro blocks according to the complexity; DCT block state determination means for classifying the image data DCT block quantized by the first quantization means into a plurality of regions according to characteristics; Second quantization step size generation means for obtaining a quantization step size according to the area of the DCT block determined by the DCT block determination means; And second quantization means for quantizing the discrete cosine-converted image data according to the quantization step size of the second quantization step size generating means.

In accordance with another aspect of the present invention, there is provided a compression encoding method, comprising: a first process of generating a first quantization step size in units of slices according to a complexity of input image data; Dividing the quantized image data DCT block into a plurality of regions according to image characteristics by DCT the input image data; And a third process of performing quantization according to the number of DCTs of each region in the second process by a second quantization step size obtained by subtracting the first quantization step size of the first process.

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

2 is a block diagram illustrating a compression encoding apparatus according to the present invention, and includes a preprocessor 210, a complexity calculator 212, a DCT generator 214, a first MQUANT generator 215, a first quantizer 216, and a DCT. Block state determiner 218, target bit calculator 220, second quantizer 222, second MQUANT generator 224, VLC 226, adder 227, subtractor 228, MQUANT adder 230 It is composed of

3 is a 4: 2: 0 image format diagram of a conventional macro block, and is composed of four luminance components (Y1, Y2, Y3, and Y4) and two color difference components (Cb and Cr). do.

4 is a typical slice format diagram of one frame.

5A is a low DCT block coefficient format diagram, FIG. 5B is a horizontal edge DCT block coefficient format diagram, FIG. 5C is a vertical edge DCT block coefficient format diagram, FIG. 5D is a diagram of a diagonal edge DCT block coefficient format, and FIG. 5E is a diagram of a high DCT block coefficient format.

As shown in FIG. 2, first, the preprocessor 210 band-limits the luminance signals Y1, Y2, Y3, and Y4 for the analog-digital converted 4: 2: 0 image data shown in FIG. Low pass filtering In this case, low pass filtering is used for band limitation in advance when fixed-length encoding is performed at a desired bit rate. The complexity calculator 212 calculates the complexity of the image data of one frame input from the preprocessor 210 in macroblock units. The complexity at this time is calculated as (pixel-average value) ² by obtaining an average value of data of one DCT block with respect to data on a spatial domain. The first MQUANT generation unit 215 generates the first MQUANT, which is an initial quantization step size to be applied in the slice unit shown in FIG. 4 based on the complexity calculated by the complexity calculator 212. The first MQUANT of the first MQUANT generating unit 215 is obtained in a slice unit complexity and is applied as a quantization step size within the same slice when quantizing each macro block unit in the first quantizer 216. That is, 1 MQUANT is the number of slices × (slice mean_act / mean_act) + α, mean_act is the total complexity for one frame, slice mean_act is the slice complexity for one slice, and α is an experimental value for the optimal value of the first MQUANT value. Α = 2.5. Next, the DCT device 214 applies the equation of TM5 of MPEG-2, and discrete cosine transforms the image data after being prefiltered from the preprocessor 210. In other words, the DCT converts the image data in the spatial domain into the frequency domain and uses the zig-zag method proposed by TM5. The first quantizer 216 performs quantization on the image data DCT obtained from the DCT generator 214 according to the initial MQUANT obtained by the first MQUANT generator 215. The quantizer at this time is also proposed by TM5, and quantization is performed by the same MQUANT for all image data. The DCT block state determination unit 218 determines the characteristics of the low, edge, and high for each DCT block for the macro block unit output from the first quantizer 216. . In other words, the image quality is visually improved by assigning weights to MQUANTs by dividing them into simple areas, complex areas, and edged areas. 5A shows the number of DCT coefficients used when determining a low DCT block. At this time, the DCT coefficients are added to the coefficients of 1,2,3,4,5,6,7,8,9 (hatched fractions) in a zigzag order. 5B shows the number of DCT coefficients used when determining the vertical edge DCT block. At this time, the DCT coefficients are summed in the zigzag order with respect to the coefficients of 1,2,4,5,6,7,14,15,27 (hatched fractions). 5C shows the number of DCT coefficients used when determining a horizontal edge DCT block. The DCT coefficients at this time are summed in the zigzag order with respect to the coefficients of 1,2,3,4,8,9,10,20,21 (hatched fractions). 5D shows the number of DCT coefficients used when determining a diagonal edge DCT block. At this time, the DCT coefficients are summed in a zigzag order with respect to the coefficients of 1,2,4,12,24,39,51,58,63 (hatched fractions). 5E shows the number of DCT coefficients used when determining a high DCT block. The DCT coefficients at this time are added to the coefficients in the zig-zag order. The DCT block determiner 218 is represented by a DCT block of low, edge, and high for the sum of the coefficients obtained by the respective DCT coefficients. At this time, the sum of the low coefficients, the sum of the three edge (horizontal vertical, diagonal) coefficients and the sum of the high coefficients divided by two are compared to determine the characteristics of the DCT block for the case having the largest value. As shown in FIG. 3, one macroblock is composed of four luminance (Y1, Y2, Y3, Y4) DCT blocks, and thus, four luminance DCT blocks are obtained.

The target bit calculator 220 allocates a target bit according to the complexity obtained by the complexity calculator 212. The target bit calculator 220 targets a given bit rate per frame according to the ratio of the complexity of each macroblock unit and the complexity of the entire frame. Allocates bits. In other words, the target bit = min_activity [mb] / mean_act × (bit_rate / 30), where min_activity [mb] is the complexity of the smallest DCT block in the macroblock, and the allocated bits are in macroblock units. Is a target bit generated by quantized by VLC and variable length coding (VLC).

The second MQUANT generator 224 obtains the second MQUANT with respect to the number of low, edge, and high DCT blocks obtained from the DCT block state determiner 218 and the first MQUANT generated by the first MQUANT generator 215 as follows. That is, 2MQUANT = ((number of low DCT blocks / 4) x 1MQUANT)-((number of edge DCT blocks / 2) x 1MQUANT) + (number of high DCT blocks / 2) x 1MQUANT.

The second quantizer 222 quantizes the DCT image data outputted from the DCT 214 using the second MQUANT. The quantizer used here is defined in TM5. The VLC device 226 performs variable length coding on the DC and AC coefficients quantized through the second quantizer 222, and performs differential pulse code modulation (DPCM) on the DC. After this, variable length coding is performed. The data output from the VLC machine 228 is an amount of actual bits generated by the encoder. The subtractor 228 subtracts the bit output from the VLC machine 226 and the target bit in macroblock units of the target bit calculator 220 to generate a surplus bit. This excess bit is added to the target bit calculator 220 and added to the target bit of the next macroblock. The MQUANT adder 230 adds or subtracts the MQUANT value according to the surplus bits generated by the subtractor 228. Table 1 shows an example for MQUANT to be added or subtracted.

Surplus bits Acceleration MQUANT Surplus bit 0 -One Surplus bit 50 +1 50 Surplus Bits 100 +2 100 Surplus Bits 200 +3

In addition, the MQUANT deceleration value shown in Table 1 generated by the MQUANT decelerator 230 and the second MQUANT generated by the first MQUANT generator 224 are added by the adder 227 to actually use the quantization step size used in the second quantizer 224. Used as MQUANT.

6 is a flowchart of a compression encoding method according to the present invention.

Operation 610 is a preprocessing process for band-limiting the input image data. In step 620, a complexity is calculated according to the input image. In operation 630, the first quantization step size MQUANT may be calculated by calculating a complexity according to the input image. Operation 640 is a process of first quantizing the image data using the first quantization step size calculated in operation 630. The process 650 is a process of dividing the regions of the low, edge, and high in order to assign weights to the characteristics of the quantized image data DCT block in step 640. In step 660, the second quantization step size is obtained by calculating the number of DCTs in the low, edge, and high states and the first quantization step size (MQUANT) of the DCT blocks. Operation 670 is a process of second quantization of the image data by the second quantization step size obtained in operation 660. In step 666, the target bit is calculated in macroblock units according to the complexity calculated in step 620. Step 680 is a process of obtaining a difference between a bit generated in the encoding apparatus and a target bit calculated in step 666. In step 690, the second quantization step is performed in the next macro block by the quantization step size obtained by adding the second quantization step size in step 660 to the quantization step size of the surplus bits in step 666.

As described above, according to the present invention, since a feedback loop does not exist, buffer control is not required, and an intra frame coding apparatus does not require an apparatus for inter frame, thereby reducing hardware size and improving intra frame coding. There is an advantage that the frame-by-frame editing is easy.

Claims (11)

In the compression coding apparatus, First quantization means for quantizing by a first quantization step size calculated with a complexity for a slice unit of input image data; Target bit calculating means for calculating a target bit in units of macro blocks according to the complexity; DCT block state determination means for classifying the image data DCT block quantized by the first quantization means into a plurality of regions according to characteristics; Second quantization step size generating means for obtaining a second quantization step size obtained by subtracting the first quantization step size according to the area of the DCT block determined by the DCT block determining means; And second quantization means for quantizing the discrete cosine transformed video data according to the quantization step size of the second quantization step size generating means. The compression encoding apparatus according to claim 1, further comprising preprocessing means for prefiltering input image data to define an image band. 2. The apparatus according to claim 1, further comprising: quantization step size adding / decreasing means for adding or subtracting a first quantization step size in accordance with a surplus bit that is a difference between all generated bits and a target bit of the target bit calculating means; And adding means for setting a quantization step size of said second quantization unit by adding a quantization step size decrement value of said quantization step size decrement means and a second quantization step size generated by said second quantization step size generation means. Compression encoding device. 4. The compression coding apparatus according to claim 3, wherein the surplus bits of said quantization step size adding / reducing means are used as target bits of a next macro block. The method of claim 1, wherein the first quantization step size (MQUANT) of the first quantization means is the number of slices × (slice mean_act / mean_act) + α, mean_act is the overall complexity for one frame, slice mean_act is for one slice Compression encoding apparatus characterized in that the slice complexity, α is an experimental value for the optimum value of the first MQUANT value. The target bit of the target bit calculating means is calculated by min_activity [mb] / mean_act × (bit_rate / 30), where min_activity [mb] is the complexity of the smallest DCT block in the macroblock. Compression coding apparatus characterized by the above-mentioned. The compression encoding apparatus according to claim 1, wherein the DCT block state determining means divides the low DCT block, the edge DCT block, and the high DCT block in units of DCT blocks. The compression encoding apparatus of claim 7, wherein the edge DCT block is a horizontal, vertical, or diagonal edge block. The method of claim 1, wherein the quantization step size of the second quantization step size generating means is ((number of low DCT blocks / 4) x first quantization step size)-((number of edge DCT blocks / 2) x first quantization step Size) + (high DCT block number / 2) x The first quantization step size is calculated. In the compression coding method, Generating a first quantization step size in slice units according to the complexity of the input image data; Dividing the quantized image data DCT block into a plurality of regions according to image characteristics by DCT the input image data; And a third process of performing quantization at a second quantization step size obtained by subtracting the first quantization step size of the first process according to the number of DCTs in each region in the second process. The method of claim 1, further comprising: calculating a target bit in macroblock units according to the complexity; Obtaining an added or subtracted value of the quantization step size according to a surplus bit, which is a difference between the bits generated by the total quantization and the target bit calculated in the process; And performing a second quantization by adding or subtracting the second quantization step size of the third process according to the value of the quantization step size obtained in the above process.