KR20080079934A - Method for inverse quantization of compressed and encoded video data stream and apparatus thereof - Google Patents

Abstract

A method for inverse quantization of a compressed/encoded video data stream and an apparatus thereof are provided to remove four addition processes from the whole process of the inverse quantization, thereby increasing the whole process speed and reducing the embodiment area of hardware. A method for inverse quantization of a compressed/encoded video data stream comprises the following steps of: shifting a frequency-converted or quantized coefficient value left or inverting the coefficient value to multiplex the coefficient value(610); adding a carry-in value to the multiplexed coefficient value and setting the lowest bit value(620); multiplying the set coefficient value by a quantization scale value(630); multiplying a multiplication result value by a quantization weight(640); and truncating the decimal place of a value that the quantization weight is multiplied and adding an added value determined according to whether the number without truncated decimal place is a positive number or a negative number or whether the number without truncated decimal place is a positive even number or a negative odd number to the remaining integral place of the truncated value(650).

Description

Inverse quantization method of compressed and encoded video data stream and apparatus

1 is a functional block diagram illustrating an operation process of a dequantization apparatus according to the prior art.

2 is a flowchart illustrating a dequantization method according to the prior art.

3 is a functional block diagram for performing a preprocessing process of inverse quantization of FIG.

4 is a functional block diagram for performing a post-processing process of inverse quantization of FIG.

FIG. 5 is a diagram illustrating the number of cases for the fraction decision of FIG. 4.

6 is a flowchart illustrating an inverse quantization method according to an embodiment of the present invention.

7 is a functional block diagram illustrating a preprocessing process of inverse quantization according to an embodiment of the present invention.

8 is a functional block diagram illustrating a post-processing process of inverse quantization according to an embodiment of the present invention.

FIG. 9 is a diagram illustrating the number of cases of the operation optimization concept of operation 820 of FIG. 8.

10 is a functional block diagram illustrating an inverse quantization apparatus according to another embodiment of the present invention.

The present invention relates to a method and apparatus for inverse quantization of a compressed coded video data stream, and more particularly to shift left or inversion with respect to frequency transformed and quantized coefficient values and carry- Inverse quantization method that adds carry-in value, sets least significant bit value, and shortens operation of coefficient value and sign mode conversion including sign conversion, mismatch control, and round step after inverse quantization And to the apparatus.

In the compression encoding process, quantization refers to a process of dividing an orthogonally transformed coding component into quantization steps and approximating the result to an integer representative value. When the quantization process is applied, the orthogonal transform component can be represented by a smaller integer value, and encoding efficiency can be increased because encoding can be performed with a smaller number of bits than encoding a component that has not undergone quantization. After the quantization process is performed by the encoder, the process of returning the input coefficients received by the decoder to the orthogonal transformed component values is called inverse quantization.

The inverse quantization process is a process of obtaining a coefficient value that is passed through an inverse discrete cosine transform (IDCT) step by processing a reconstructed coefficient through an entropy decoding step, and in decoding a compressed coded video signal such as digital TV broadcasting. It is an indispensable process.

The prior art related to the inverse quantization method is described in US Patent No. 5784011, "Multiplier Circuit for Performing Inverse Quantization Arithmetic" and US Pat. No. 6301304, "Architecture and Method for Inverse Quantization of Discrete Cosine Transform Coefficients in MPEG Decoders" You can take a look.

1 is a functional block diagram illustrating an operation process of a dequantization apparatus according to the prior art.

In general, in MPEG, which is a video signal compression coding standard, inverse quantization is performed by receiving input coefficients (coefficient, coeff [m] [n], 110) that have undergone entropy decoding as shown in FIG. 1. The de-quantization parts 120 to 140 according to DC, intra AC, and inter DC / AC, a multiplexer 150 for selecting these results according to a coding mode of coefficients, and a dequantization result. Each consists of an MPEG-1 mismatch control section 160, an MPEG-2 mismatch control section 170, and a multiplexer 180 that selects two results.

As will be described later, in the intra DC mode, the input coefficient (coeff [m] [n], hereinafter referred to as “C”) is immediately mismatched according to the input coefficient mode (intra DC, intra AC, inter DC / AC). Entering the control step, the operation is performed like the equation in block 140 in the intra AC mode, and the equation in the block 130 in the inter DC / AC mode. In the formulas shown in blocks 130 and 140, sign [m] [n] has a value of -1 if negative and +1 if positive, according to the sign of the coefficient (coeff [m] [n]). It is a rounding operation that performs round-to-zero. The quantization scale value q_scale, the inter mode quantization matrix Nonnintra_Qmat [m] [n], and the intra mode quantization matrix Intra_Qmat [m] [n] are selected according to the entropy decoded information and used in the process of inverse quantization. .

2 is a flowchart illustrating a dequantization method according to the prior art.

In order to perform the MPEG dequantization described in FIG. 1, the prior art implements dequantization in the same order as in FIG. 2. In the inverse quantization method, steps 200 to 230 may be largely preprocessed, and steps 240 to 290 may be divided into postprocessing.

Looking at each step, the preprocessing step is to calculate the coefficient value as (2C + 1), (2C-1), 2C, C according to the encoding mode of the input coefficient (C) (200), the result is an unsigned mode converting to an unsigned mode (210), multiplying the quantized scale value (q_scale) (unsigned multiplier, 220), and converting it back to a signed mode (signed mode) so that it becomes the original sign (230). do.

Subsequent post-processing is as follows. Converting the preprocessing result back to unsigned mode (240), multiplying the quantized weight (q_weight) (unsigned multiplier, 250), and signing mode to convert the result back to its original sign In the conversion step 260, in the case of an MPEG-1 image, determining a fraction part for MPEG-1 mismatch control (270), rounding by adding a fractional value; Step 280 of performing MPEG-1 mismatch control, and step 290 of performing MPEG-2 mismatch control in case of an MPEG-2 image.

Looking at the inverse quantization process of the prior art with reference to Figure 2, in the pre-processing step, first extract the sign information of the coefficient, and then calculate (2C + 1), (2C-1), 2C, C according to the encoding mode In case of a negative number, 2's complement is performed to make an unsigned number (210). The result is multiplied by a quantization scale value (q_scale), which is a macroblock-processed value (220), and then changed to a signed number according to the original sign before passing to the post-processing step (230). In the next post-processing step, the sign information is extracted from the signed number passed in the pre-processing step, and then changed back to an unsigned number (240), and the quantization weight (q_weight), which is a processed value in picture units, is changed. After multiplying to obtain a result (250), it is converted back to a signed number according to the sign information (260). In the case of finally MPEG-1 for this result value, a fraction for rounding and MPEG-1 mismatch control is determined (270), and the determined fraction is converted to a previous step. The result is added to the result of 260 (280). In the case of MPEG-2, MPEG-2 mismatch control is performed after rounding (290).

Here, the prior art requires an operation for calculating a coefficient and changing a sign according to a mode, and a fraction decision operation for simultaneously performing round and MPEG-1 mismatch control. That is, in the preprocessing part, (2C + 1), (2C-1), 2C, and C generating part require two adders by 2C-1 and 2C + 1 operation, and again, the unsigned value ( To convert to unsigned number, a 2's complement is required, which also requires an adder. Next, in the post-processing part, an adder is required because two's complement processing is required in the process of converting to a signed value. In order to simultaneously perform round and MPEG-1 mismatch control, a fraction decision is made and an operation is performed in the next adder.

3 is a functional block diagram for performing a preprocessing process of inverse quantization of FIG.

Referring to FIG. 3, two operations of performing (2C + 1), (2C-1), 2C, and C to perform a preprocessing process and converting the mux part 310 to an unsigned mode are performed. A multiplexing part 320 that multiplies the 2's complement part with the multiplexing part and the quantization scale value (q_scale) and selects a multiplication result, and finally the complementing part and multiplexing part of 2 that converts to signed mode ( 330).

4 is a functional block diagram of performing a post-processing process of inverse quantization of FIG. 2.

Referring to FIG. 4, a multiplier portion of 2, which converts to an unsigned number, a multiplier portion, and a multiplier portion 410 that multiplies quantization weights (q_weight), a 2's complement that converts to a signed number. A part and multiplex part 420, a fraction decision part and an adder part 430 that adds the result of the signed value with the determined fractional value and the MPEG-2 depending on whether it is an MPEG-2 video signal or not. And a multiplexing portion 440 that performs and selects mismatch control.

FIG. 5 is a diagram illustrating the number of cases for the fraction decision of FIG. 4.

In each case of 520 to 560, the case where the added number is determined according to a condition in the fraction decision part of FIG. 4 is classified in detail.

When a signed number input is a signed number input, five bits from the least significant bit represent a decimal value, and 23 more bits represent an integer value (510), and (i) an integer value of zero In the case of a small negative value, +31 is added to be integer based on 0 (round-to-zero, RTZ), and (ii) in the case of positive even value, -32 is added to MPEG-1. Perform mismatch control. (iii) In case of negative even value, +63 is added to perform RTZ and MPEG-1 mismatch control. (iv) In case of negative odd value, +31 is added to add RTZ. (V) add +0 for the other cases.

However, this prior art has three major problems.

First, an adder is needed to generate (2C + 1), (2C-1), 2C, and C in the preprocessing portion, and an adder is needed to convert it back to unsigned mode. Second, a conversion to signed portion is required before passing the result of the operation in the preprocess to the original sign, which is then converted to unsigned in order to perform multiplication on an unsigned value. There is an unnecessary step. Third, in the post-processing step, a conversion to signed step is performed to replace the result of an unsigned number with the original sign, and after the fractional end decision is performed to perform round and MPEG-1 mismatch control, An adder is needed to add the fractional steps and the result of the previous step.

Therefore, the prior art requires a plurality of adders in each operation as described above, converts the number entered as a signed number at each step into an unsigned number, and converts it back to its original state. signed, so this sign conversion occurs redundantly. In addition, since the round-to-zero operation is performed in the signed mode and the MPEG-1 mismatch control is performed at the same time, the fraction decision process is complicated, and when the fraction is added, We need an adder that is as large as the bit width that contains the part. Therefore, the prior art requires a large number of adders, thus increasing the area when the device is implemented, and also requires complicated control during driving, thereby increasing the total processing time of the operation.

Accordingly, the present invention has been devised to solve the above problems, and a technical problem to be achieved by the present invention is to provide a method and apparatus for optimizing arithmetic operations such as code conversion, rounding, and mismatch control during inverse quantization.

According to an aspect of the present invention, there is provided a method of inverse quantization of a compression-coded image data stream, the method comprising: performing multiplexing by shifting left or inversion of frequency transform and quantized coefficient values; Adding a carry-in value to a value, setting a least significant bit value, and multiplying the set coefficient value by a quantization scale value.

And multiplying the multiplied result by the quantization weight, truncating the fractional stage of the multiplied quantization weight value, and whether the result is positive or negative with respect to the remaining integer stage of the truncated value. It is preferable to further include the step of adding the added value determined according to whether or not to the water purification stage.

Multiplying the quantization scale value and the quantization weight value is preferably performed without sign conversion.

According to another aspect of the present invention, an object of the present invention is to provide a shifter and inversion of a frequency transformed and quantized coefficient value in a dequantization apparatus of a compression coded video data stream. A multiplexer multiplexed according to whether an inverter, the shifter and the inverter are performed, an adder for adding a carry-in value to a coefficient value multiplexed in the multiplexer, and a bit setting unit for setting a least significant bit value and the adder And a multiplier for multiplying the coefficient value processed by the bit setting unit with the quantization scale value.

The present invention further comprises a computer readable recording medium having recorded thereon a program for implementing the dequantization method.

Hereinafter, with reference to the accompanying drawings will be described a preferred embodiment of the present invention.

6 is a flowchart illustrating an inverse quantization method according to an embodiment of the present invention.

Steps 610 to 630 are the pretreatment parts and the remaining steps 640 to 650 are the post-treatment parts.

In general, a step 610 of shift left or inversion of the frequency transform and quantized coefficient values and a carry-in value for the multiplexed coefficient values are added and the lowest values are added. A step 620 of setting a bit value, a step 630 of multiplying the set coefficient value by a quantization scale value, a step 640 of multiplying the quantization weight, and a truncation of the fractional end of the value of the quantization weight multiplied, Comprising a step 650 to add the added value determined in accordance with the positive or negative and even / odd for the remaining integer stage of the truncated value to the integer stage.

First of all, the signed / unsigned conversion, which was performed unnecessarily between the pre-processing part and the post-processing part in the prior art, is eliminated. This unnecessary conversion can be omitted by continuing with an intermediate operation in the unsigned conversion state and performing a signed conversion at the final output stage. Hereinafter, the operation process will be described in detail.

7 is a functional block diagram illustrating a preprocessing process of inverse quantization according to an embodiment of the present invention.

First, the process of calculating the coefficient value (C) with (2C + 1), (2C-1), 2C, and C and unsigned conversion is performed simultaneously (710), and the coding mode of the coefficient is (i). In case of intra DC, C and -C operations are required. (Ii) In case of intra AC, -2C and 2C operations are required depending on the negative and positive values. (Iii) In case of negative coefficients in inter DC / AC, Since (2C-1) needs to be converted to a positive number,-(2C-1) = -2C + 1 is required, and for positive coefficients, (2C + 1) is required. As a result, a total of six results of (C, -C, -2C, 2C, -2C + 1, 2C + 1) can be obtained according to the conditions.

These six operations can be optimized for each case. First, assume that the coefficient C is a binary number consisting of {w, w, x, y, z} bits, and w is a sign bit. The underlined uppercase letters ( W , X , Y , Z ) are also defined as inversion values for w, x, y, and z. Thus, the six calculation results (C, -C, -2C, 2C, -2C + 1, 2C + 1) can be calculated as follows.

(i) In the case of C, the input coefficient C may be selected and output as it is.

(ii) In the case of -C, it is enough to invert the input C and add 1 to the least significant bit (LSB). To transform it again, invert C except LSB, leave the input z intact, and add Z to the second LSB. This is consistent with the other cases (iii) ~ (vi) to facilitate the hardware implementation.

(iii) In the case of -2C, after performing arithmetic shift left 1 bit on the input C, the part except the LSB is inverted and 1 is added to the second LSB. Thus, LSB becomes zero.

(iv) In the case of 2C, input bit shifts 1 bit to the left. Thus, LSB becomes zero.

(v) In the case of -2C + 1, +1 may be performed on the result of (iii). Therefore, the LSB of 0 in (iii) should be replaced with 1.

(vi) In the case of 2C + 1, +1 may be performed on the result of (iv). Therefore, the LSB of 0 in (iv) should be replaced with 1.

Considering all six cases above, the LSB is selected from values of z, 1, and 0. The carry-in value input to the second LSB is selected from values of Z , 1, and 0. Bits other than the LSB of C may be determined by a combination of arithmetic shfit left 1 bit and inversion. Next, the LSB decision and the addition operation for the second LSB occur in a carry-in decision / LSB decision circuit (720), and the result value is multiplied by the quantization scale value (Q_scale) and finally output (730). . At this time, in the case of intra DC, the multiplication process is bypassed.

8 is a functional block diagram illustrating a post-processing process of inverse quantization according to an embodiment of the present invention.

Referring to FIG. 8, first, a quantization weight (q_weight) is multiplied by an input value (810). Next, an operation according to MPEG-1 mismatch and an added number for sign conversion are determined for the multiplication result, and the result is added by the adder (820). If the processed data is an MPEG-2 video signal, an MPEG-2 mismatch operation may be additionally performed without performing MPEG-1 mismatch (830).

A detailed decision process for MPEG-1 mismatch and sign conversion in the post-processing process can be seen in FIG. 9. 9 is a diagram illustrating the number of cases of the operation optimization concept of step 820 of FIG. 8.

First, in the prior art, since the fractional truncation (RTZ) was performed in the signed mode, it was necessary to consider the fractional part. In the present invention, the RTZ is performed in the unsigned mode. Therefore, the fractional end is truncated (910). Next, the remaining operations are oddification and singed number corresponding to MPEG-1 mismatch, which can be integrated and implemented according to the following rules.

(i) That is, the input integer part is inverted if it is negative and not inverted if it is positive. (ii) The added number added to the input integer part is MPEG-1 mismatch control and sign conversion ( Signed conversion is simultaneously performed (920), and in the case of MPEG-2, only sign conversion is performed (930).

Looking specifically at each case,

(case 1) If the input is negative and even even, perform two's complement and add +1 for oddization. Thus the integer input is inverted and +2 is added by adding +1 for two's complement and +1 for oddification.

(case 2) In case of negative and negative odds, only two's complement is performed.

(case 3) If it is positive and positive even, the input is not inverted and only -1 is added for oddification. The two's complement of -1 needs to set all input bits to 1.

(case 4) In the case of a positive number and a positive odd value, 0 is added since two's complement operation and oddification are not necessary.

(case 5) Since it is the case of MPEG-2, in the case of a negative input, the input is inverted and +1 is added for the two's complement conversion.

(case 6) Since it is the case of MPEG-2, in the case of positive input, since no operation is required, 0 is added.

10 is a functional block diagram illustrating an inverse quantization apparatus according to another embodiment of the present invention.

Looking at the overall configuration, a shifter 1010 shifting left and frequency shifted quantized coefficient values and an inverter 1020 and a multiplexer multiplexed according to whether the shifter and the inverter are performed ( 1030), an adder 1040 for adding a carry-in value to the coefficient value multiplexed in the multiplexer, a bit setting unit 1050 for setting the least significant bit value, and a coefficient processed in the adder and bit setting unit. Multiplier 1060 that multiplies the value by the quantization scale value.

In addition, the multiplier 1060 may multiply quantization weights for post-processing, and adds a truncation unit 1070 which cuts a fractional end for a round, and an added value for MPEG-1 mismatch control and code conversion. The value determiner 1080 and an adder 1090 for adding the same may be configured.

The dequantization method of a compression-encoded video data stream according to the present invention can also be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include magnetic storage media such as ROM, RAM, magnetic tape and floppy disks, and optical read media such as CD-ROMs and optical data storage devices. It also includes the implementation in the form of (transmission through). The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will understand that the present invention may be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

As described above, according to the inverse quantization method of a compression-encoded video data stream, the following improvement effect is obtained.

First, in order to generate (2C + 1), (2C-1), 2C, and C in the preprocessing step, and to convert to unsigned, one process can be processed by one adder. The process is optimized to integrate.

Second, by eliminating unnecessary steps of conversion to signed and conversion to unsigned before multiplication in the post-processing step. The two adders are removed.

Third, in the post-processing step, the round process is changed to a truncation process in an unsigned mode, thereby eliminating the need for the fractional part. In addition, the MPEG-1 mismatch control and the conversion to signed step are optimized to be implemented as one adder.

Therefore, the above improvement increases the overall processing speed by eliminating four addition steps in the entire stage of dequantization, and also reduces the implementation area in hardware implementation.

Claims (5)

In the dequantization method of a compression-coded video data stream, Shift left or inversion multiplexing the frequency transform and the quantized coefficient values; Adding a carry-in value to the multiplexed coefficient value and setting a least significant bit value; And multiplying the set coefficient value by a quantization scale value. The method of claim 1, Multiplying a quantization weight by the multiplied result; Truncating the fractional end of the value multiplied by the quantization weight; And adding an additional value determined according to whether the coefficient value is positive / negative and even / odd with respect to the remaining integer stage of the truncated value. The method of claim 2, And multiplying the quantization scale value and the quantization weight value by a signal conversion without sign conversion. In the inverse quantization apparatus of a compression-coded video data stream, A shifter for shifting left and frequency shifting the quantized coefficient values and an inverter for inversion; A multiplexer multiplexed according to whether the shifter and the inverter are performed; A bit setting unit for setting an adder for adding a carry-in value to a coefficient value multiplexed in the multiplexer and a least significant bit value; And a multiplier for multiplying a coefficient value processed by the adder and the bit setting unit with a quantization scale value. A computer-readable recording medium having recorded thereon a program for implementing the dequantization method according to any one of claims 1 to 3.

Images