KR20110070687A - Encoding apparatus and method capable of multi-dimension transforming and quantization, and decoding apparatus and method - Google Patents

Abstract

PURPOSE: An encoding apparatus and method capable of multi-dimensional conversion and quantization and a decoding apparatus and method are provided to apply integer-type quantization to a difference image which is converted twice. CONSTITUTION: A first differentiator(140) removes a prediction block created by inter prediction and creates a difference image. A converting unit(151) performs the multi-dimensional conversion of the difference image. A quantization unit(153) applies integer-type quantization to the multi-dimensionally converted difference image. An entropy coding unit(157) encodes and compresses the coefficient values of the re-aligned difference image.

Description

Coding apparatus and method capable of multi-dimensional transformation and quantization, and decoding apparatus and method TECHNICAL FIELD

The present invention relates to a coding apparatus and method capable of multi-dimensional transformation and quantization, and to a decoding apparatus and method, which enables efficient video data compression and reconstruction by at least three-dimensional transformation and integer quantization.

The present invention is derived from the research conducted as part of the IT source technology development of the Korea Communications Commission [Task Management Number: 2008-F-011-01, Task name: Development of next-generation DTV core technology].

When the demand for multimedia is increasing, the technology of compressing video is very important. However, as the market for high-definition and high-resolution images has recently expanded, research for codecs with high efficiency has been conducted.

In general, the H.264 / AVC standard established by the World Standards Organization for video compression is widely used. H.264 / AVC, a video data compression standard, applies transformation of 4x4 block unit whose block size is smaller than that of the existing standard. This can reduce the difference image error, but is inefficient to reduce the spatial redundancy. In addition, in the H.264 / AVC standard, in addition to the 4x4 block unit, the 8x8 block unit that can be used in the high profile or higher can reduce the spatial redundancy while the block size is larger than that of the 4x4 block. There is a growing problem.

In addition, the quantization process is also affected by the transformation in the compression process. That is, the DCT method, which is mainly used as a transformation, performs a decimal point operation, and thus a miss match phenomenon occurs in which values between the encoder and the encoder do not match each other.

Accordingly, an object of the present invention to solve the above problems is to encode a multi-dimensional transformation and quantization in order to reduce the difference image error, and to reduce the spatial redundancy by using a feature of high spatial correlation between adjacent blocks on the image characteristics and A method, and a decoding apparatus and method.

According to an embodiment of the present invention, an encoding apparatus capable of multi-dimensional transformation and quantization includes: a transformer configured to at least two-dimensionally transform a difference image generated by removing a prediction image generated by inter prediction from an original image, at least twice; And a quantization unit applying integer quantization to the at least three-dimensional transformed difference image.

The quantization unit performs the integer quantization in units of macroblocks. When the quantization coefficient value to be currently encoded is 1, the quantization unit may determine whether to encode the macroblock in consideration of a quantization coefficient having a previously encoded quantization coefficient value of 0. have.

The quantization unit may apply the integer quantization by applying rounding or rounding after applying a quantization offset value to coefficient values corresponding to upper high frequencies among the quantization coefficient values of the difference image.

The quantization unit may apply the integer quantization by applying rounding or rounding after applying a quantization offset value to coefficient values corresponding to lower low frequencies among the quantization coefficient values of the difference image.

The quantization unit applies a quantization offset of 1 / M to coefficient values corresponding to lower M low frequencies among quantization coefficient values of the difference image, and applies coefficient values corresponding to upper N (N <M) high frequencies. The interval length Δ of the quantization coefficient may be adjusted by applying a quantization offset of 1 / N.

The quantization unit may apply a quantization offset value used in encoding an inter frame to remaining coefficient values except for upper M and lower N coefficient values among the quantization coefficient values.

The quantization unit applies a scanning method in a direction perpendicular to a horizontal direction that is a direction of the horizontal mode for a 16 × 8 horizontal mode, and scans in a direction perpendicular to a vertical direction that is a direction of the vertical mode for an 8 × 16 vertical mode. The method can be applied.

On the other hand, the encoding method capable of multi-dimensional transformation and quantization according to an embodiment of the present invention comprises the steps of: multi-dimensional transformation of the difference image generated by removing the prediction image generated by inter prediction from the original image; And an integer quantization of the multi-dimensional transformed difference image.

On the other hand, the decoding apparatus according to the present embodiment, an entropy decoding unit for extracting the integer quantized coefficient value and the motion vector by entropy decoding the input bitstream; An inverse quantizer for generating coefficient values of a difference image by applying integer inverse quantization to the extracted integer quantized coefficient values; And an inverse transform unit generating a difference image by performing multi-dimensional inverse transformation on the coefficient value of the generated difference image.

On the other hand, the decoding method according to the present embodiment, entropy decoding the input bitstream to extract an integer quantized coefficient value; Generating coefficient values of a difference image by applying integer inverse quantization to the extracted integer quantized coefficient values; And generating a difference image by multi-dimensional inverse transformation of the coefficient value of the generated difference image.

According to an embodiment of the present invention, the efficiency of three-dimensional transform is increased by quantization and encoding corresponding to three-dimensional transform and three-dimensional transform, which reduce difference image error and reduce spatial redundancy, and consequently, compress video data. It can improve performance.

In addition, according to an embodiment of the present invention, two-dimensional and four-by-one transforms, which are 4x4 transforms and 4x1 transforms, are used rather than the 4x4 transforms or 8x8 transforms used in the H.264 / AVC high profile. Using dimensional transformations together can provide more efficient compression performance.

In addition, according to an embodiment of the present invention, the threshold value is adjusted according to the three-dimensional transform and its characteristics, and the fast encoding method according to the bit reduction and integerization through the scanning method according to the mode of the three-dimensional transform and integer three-dimensional quantization Can provide.

1 is a block diagram illustrating a video data compression apparatus including an encoding apparatus capable of multidimensional transformation and quantization according to an embodiment of the present invention;
2 is a view for explaining a three-dimensional transform performed in the transform unit according to an embodiment of the present invention;
3 is an integer quantization table corresponding to a three-dimensional transform of the transform unit;
FIG. 4 is a diagram for explaining the case where the quantization offset used in the quantization unit is applied to four 4x4 blocks in an 8x8 block.
5A and 5B illustrate different scanning schemes for 16x8 mode and 8x16 mode after performing 3D transform and quantization in the transform unit and the quantizer and before entropy encoding is performed in 8x8 block units. Drawing showing an example,
6 is a flowchart illustrating a coding method capable of multidimensional transformation and quantization according to an embodiment of the present invention;
7 is a block diagram illustrating a video data reconstruction apparatus including a decoding apparatus capable of integer inverse quantization and multi-dimensional inverse transformation according to a proposed embodiment;
8 is a flowchart provided to explain a video decoding method according to an embodiment of the present invention.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In describing the present invention, when it is determined that detailed descriptions of related known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. Terminology used herein is a term used to properly express a preferred embodiment of the present invention, which may vary according to a user, an operator's intention, or a custom in the field to which the present invention belongs. Therefore, the definitions of the terms should be made based on the contents throughout the specification.

1 is a block diagram illustrating a video data compression apparatus including an encoding apparatus capable of multidimensional transformation and quantization according to an embodiment of the present invention.

Referring to FIG. 1, the video data compression apparatus 100 includes an intra predictor 110, a motion estimator 120, a motion compensator 130, a first divider 140, an encoder 150, and an inverse. A quantization and inverse transform unit 160, a second difference unit 170, a filter 180, and a frame memory 190 are included.

The intra predictor 110 performs intra prediction using a reconstructed pixel around the current frame Fn. That is, the intra predictor 110 predicts and encodes a macroblock to be encoded using spatially adjacent pixel values in order to encode a block of a current image.

The motion estimator 120 estimates a motion vector using the current frame and the past frame to be encoded, and the motion compensator 130 predicts the prediction value of the current block based on the motion vector estimated by the motion estimator 120. Create a block. The motion estimation unit 120 may estimate the motion by using various motion estimation algorithms such as block matching algorithm (BMA), phase correlation, and HSBMA.

The switch S provides the first block 140 with the prediction block output from the intra predictor 110 or the motion compensator 130. The switch 140 may select an excellent prediction among intra prediction and inter prediction.

The first branch 140 generates a difference block by subtracting the prediction block from the current block.

The encoder 150 performs an intra prediction or inter prediction encoding on the current block based on one of the intra prediction direction determined by the intra prediction unit 110 or the inter prediction predicted by the motion compensator 130 to perform a bit stream. Create To this end, the encoder 150 includes a transformer 151, a quantizer 153, a reordering unit 155, and an entropy coding unit 157.

The transform unit 151 performs Discrete cosine transform (DCT) on the difference block generated in the first branch 140. In particular, the converter 151 may perform at least two-dimensional transformation by converting the difference image generated by removing the predicted image from the original image at least twice. The original image may be a current block of the current frame, and the prediction image may be a prediction block.

In order to increase the efficiency of the transformation method, the transform unit 151 does not apply a 3D transform to an intra-predicted block, but 3D transform the inter-predicted block, that is, a difference image generated by inter prediction. Can be applied. In particular, the transform unit 151 performs an integer DCT to avoid a decimal point operation in the DCT process, which means that the three-dimensional transform in the transform unit 151 is based on an integer DCT.

The quantization unit 153 generates quantization coefficients by integer-quantizing the difference image transformed by the DCT block or the 3D transform method in macroblock units. In addition, when the transform unit 151 performs at least three-dimensional transformation of the difference image, the quantization unit 153 may apply integer quantization to the at least three-dimensional transformed difference image using a quantization offset.

For example, when inter prediction is selected and the transform unit 151 applies a three-dimensional transform that transforms the difference image twice, the quantization unit 153 may obtain coefficients obtained by the three-dimensional transform to correspond to the three-dimensional transform. Quantize the value to an integer, which improves compression performance. The converter 151 performs an integer type DCT to avoid a decimal point operation in the DCT process. That is, since the 3D transform in the transform unit 151 is based on the integer DCT, the quantization unit 153 may perform integer quantization using the quantization table of FIG. 3 or the quantization offset of FIG. 4 suitable for the integer DCT. Can be.

The quantized results are rearranged in reordering unit 155 for encoding.

The entropy coding unit 157 entropy encodes encoding information such as a quantized transform coefficient and a motion vector to generate a bitstream.

The inverse quantization and inverse transform unit 160 inversely quantizes the differential blocks quantized by the DCT and the quantization unit 153 for use in prediction of the next frame to be encoded, and inverse discrete cosine transform to restore the differential blocks before encoding. .

The second divider 170 adds the reconstructed differential block and the predictive block generated by the motion compensator 130 to reconstruct the current block before encoding. The reconstructed block or reconstructed frame is stored in the frame memory 190 via the filter 180 and used for inter prediction on the next block or the next frame.

2 is a view for explaining a three-dimensional transform performed in the transform unit according to the proposed embodiment. Referring to FIG. 2, the transform unit 151 performs two-dimensional transformation (A-> B) on a 4x4 block, which is a subblock of four 8x8 blocks included in a macroblock, and performs the same frequency among the four blocks that have been two-dimensionally transformed. Four coefficient values at the position may be collected, and one-dimensional transformation (B-> C) may be performed in units of 4 × 1 blocks to array 16 difference images. That is, according to FIG. 2, the transform unit 151 performs at least two integer transformations to effectively code the difference image.

In addition, the transform unit 151 may collect four adjacent 4x4 blocks in different ways according to a mode for 3D transformation. In particular, in the 16x8 mode, the transform unit 151 brings four 4x4 blocks in units of 16x4 blocks to increase the correlation between adjacent blocks. In addition, in the 8x16 mode, the conversion unit 151 brings four 4x4 blocks in units of 4x16 blocks. At this time, the scanning method to be described later in Figures 5a and 5b is adaptively applied to reflect the gathering form between blocks.

3 is an integer quantization table used in the quantization unit proposed according to the three-dimensional transformation of the transform unit.

When the 3D transform is applied by the transform unit 151, the dynamic range may be different from the H.264 / AVC standard. The converter 151 may convert the operating range of the kernel and the difference image error value into 15 bits in the 4x4 transform, that is, the 2D transform. Since the 3D transform performs a 1D transform in addition to the 4x4 transform, the result is to multiply the 1D kernel once more. Therefore, the operating range is converted into 17 bits. Therefore, the quantization bit values qbits used by the quantization unit 153 in the quantization process are (17 + QP / 6). [Equation 1] is an example of expressing this as an equation.

Since the 3D transformation further performs a 1D transformation in addition to the 2D transformation, as shown in FIG. 3, the MF table value and the V table value are changed, and two may exist. In FIG. 3, the vertical axis is QP (Quantization Parameter), and the horizontal axis is grouped and labeled according to the position of 4 × 4 blocks. The MF table values and the V table are tables used for quantization in H.264.

If the x and y position of a pixel in a 4x4 block is represented by coordinates (x, y), then r = 0 in the table (A ~ D) is (0,0), (0,2), (2,0) , The position of (2,2), r = 1 is the position of (1,1), (1,3), (3,1) and (3,3), and r = 2 corresponds to the rest of the 4x4 block do. The quantization unit 153 applies the tables A and C to the 4x4 blocks at the positions b0 and b2 (Fig. 4) in the 8x8 block, and the tables B and B to the 4x4 blocks at the positions b1 and b3 (Fig. 4) in the 8x8 block. D) is applied to perform integer quantization.

FIG. 4 is a diagram for explaining a case where the quantization offset used in the quantization unit is applied to four 4x4 blocks in an 8x8 block.

Since the transform unit 151 performs the transformation twice (three-dimensional and one-dimensional transformation) for the three-dimensional transformation, the quantization unit 153 has a dead zone of the coefficient value of the DC position having a high importance. By reducing the size, 0 value is less, and the coefficient value of high frequency position with less importance can increase the dead zone size so that 0 value is much.

In detail, the quantization unit 153 applies integer quantization by rounding up or down after applying a quantization offset value to coefficient values corresponding to the lower M low frequency positions among the quantization coefficient values of the difference image. The dead zone can be reduced so that zero values are produced. The quantization unit 153 may apply 1 / M as a quantization offset to coefficient values corresponding to the lower M low frequency positions. The dead zone is a length of a section including a coefficient value 0, and the quantization unit 153 may adjust the section length by applying a quantization offset.

In addition, the quantization unit 153 applies a quantization offset value to coefficient values corresponding to the top N high frequency positions among the quantization quantization coefficient values of the difference image multi-dimensionally transformed by the transform unit 151, and then rounds or rounds them down. Integer quantization is applied, thus increasing the dead zone so that a large number of zero values can be obtained. The quantization unit 153 may apply 1 / N to the upper N coefficient values as a quantization offset. At this time, M = 3 and N = 9.

In addition, the quantization unit 153 may apply the quantization offset value used when encoding the inter frame to the remaining coefficient values except for the upper M and lower N coefficient values among the quantization coefficient values.

For example, the quantization offset used in the quantization unit 153 in inter encoding of an algorithm such as H.264 / AVC may be applied according to a 3D transform characteristic. The quantization unit 153 applies a quantization offset of three coefficient values of DC digits generated by two-dimensional transformation after the two-dimensional transformation by the transformation unit 151 to 1/3, and outputs by one-dimensional transformation. A quantization offset of nine coefficients of high frequency may be applied as 1/9. The quantization unit 153 may apply 1/6, which is a quantization offset value used to encode an inter frame in the H.264 / AVC standard, for coefficient values at other positions.

In the case of FIG. 4, as described with reference to FIG. 2, the transform unit 151 performs two-dimensional transformation in units of 4 × 4 blocks, and performs one-dimensional transformation of coefficient values of the same frequency position among four adjacent 4 × 4 blocks. In addition, the quantization unit 153 applies the quantization offset of the coefficient values of the DC positions 1,2 and 3 to 1/3 while quantizing the coefficient values output from (C), and performs the one-dimensional conversion to obtain the highest frequency position. The quantization offset of the coefficient values of 5, 8, 9, 11, 12, 13, 14, and 15 is applied to 1/9. The quantization unit 153 applies 1/6, which is a quantization offset value used to encode an inter frame, in the H.264 / AVC standard for coefficient values at other positions.

5A and 5B are diagrams illustrating an example in which the quantization unit or the rearrangement unit applies the scanning method differently to the 16x8 mode and the 8x16 mode before performing the 3D transform and quantization in the transform unit and the quantization unit and performing entropy encoding on an 8x8 block basis. .

The transform unit 151 two-dimensionally converts the difference image into 4x4 block units, generates four adjacent 4x4 blocks in the form of 16x4 and 4x16 blocks in the 16x8 and 8x16 modes, and then the rearrangement unit 155 is the same as the four. Since the coefficient values of the frequency positions are collected and encoded, the scanning method can also follow the directionality of the transform encoding according to the mode.

In addition, the quantization unit 153 applies the scanning method in a direction perpendicular to the horizontal direction in the horizontal mode for the 16x8 horizontal mode, and perpendicular to the vertical direction in the vertical mode in the 8x16 vertical mode. The scanning method can be applied in the direction.

For example, referring to FIG. 5A, the transform unit 151 and the quantization unit 153 collect three adjacent 4x4 blocks that match the characteristics of the 16x8 horizontal mode, and then three-dimensional transform and quantize them. Reference numeral 155 takes the frequency coefficient values in the order written in the 16x4 block and arranges them in the zigzag scanning method in the 8x8 block.

Referring to FIG. 5B, the transform unit 151 and the quantization unit 153 collect three adjacent 4x4 blocks that match the characteristics of the 8x16 vertical mode, three-dimensional transform and quantize them, and then the reordering unit 155. ) Takes the frequency coefficient values in the order written in the 4x16 block and arranges them in the zigzag scanning method in the 8x8 block.

In addition, the quantization unit 153 performs integer quantization in units of macroblocks as described above, and when the quantization coefficient value to be currently encoded is 1, the macroblock is considered in consideration of a quantization coefficient having a previously encoded quantization coefficient value of 0. Can be encoded.

Hereinafter, referring to [Equation 2] and [Equation 3], the quantized coefficient value is 1 after transforming the difference image in 8x8 macroblock units, and when the number of quantized coefficient values is 1 is very small, An example of determining whether entropy is encoded will be described.

[Equation 2] and [Equation 3] show the case where the coefficient value obtained by transforming and quantizing the difference image in 8 × 8 block units is 1, and the number thereof is considerably small, and is transmitted without encoding the difference image. c (ii, jj) is 0, 1, 2, 3,... , Coeff (ii, jj) is a coefficient value that has been three-dimensionally transformed and quantized, ii is an 8x8 block index, and jj is an 8-dimensional 8x8 block coordinate. rr denotes the number of continuous counting of the coefficient value 0 encoded before, when the quantization coefficient value to be currently encoded is 1.

In Equation 2, when the quantized coefficient value is 1 or more, the corresponding cost c (ii, jj) is infinity, thus deviating from the condition. In this case, the entropy coding unit 157 encodes and transmits the quantized coefficient value of the 8x8 macroblock. When the quantized coefficient value is 1, the quantization unit 153 assigns 3, 2, 1, or 0 values as costs according to the rr value.

Referring to Equation 3, if the result of adding the cost c (ii, jj) from the 8x8 macroblock is less than or equal to the threshold value 2, the quantization unit 153 determines the quantized coefficients of the entire 8x8 macroblock. The value is determined to be 0, and the entropy coding unit 157 does not encode the coefficient value of the macroblock and therefore does not transmit it. As a result, the bit rate compared to the same PSNR can be improved.

According to the above-described embodiment of the present invention, when the inter prediction is selected, the transform unit 151 applies a three-dimensional transform that transforms the difference image twice, and the quantization unit 153 applies a three-dimensional transform to correspond to the three-dimensional transform. Compression performance can be improved by quantizing the coefficient values obtained by the dimensional transform into integers.

6 is a flowchart illustrating an encoding method capable of multidimensional transformation and quantization according to an embodiment of the present invention.

6 illustrates an example of three-dimensional transformation and quantization when four 4x4 submacroblocks are included in an 8x8 macroblock.

In operation 610, the first branch 140 may generate a difference image by removing a prediction block generated by inter prediction from a current block of the current image. Inter prediction is a prediction mode performed by the motion estimator 120 and the motion compensator 130.

In operation 620, the transform unit 151 may 2-dimensionally convert the generated difference image. For example, the transform unit 151 may generate coefficient values by two-dimensional transforming (A-> B) four 4 × 4 blocks included in an 8 × 8 block as shown in FIG. 2. have.

In operation 630, the transformation unit 151 may perform one-dimensional transformation on the two-dimensional transformed difference image and perform at least two conversions. For example, the transform unit 151 may collect four coefficient values located at the same frequency position among four blocks that have been two-dimensionally transformed, and may additionally perform one-dimensional transform in units of 4 × 1 blocks, as shown in FIG. 2.

In step 640, the quantization unit 153 performs integer quantization using the quantization offset with respect to the coefficient value of the 3D transformed difference image. In detail, in operation 640, the quantization unit 153 applies integer quantization by rounding up or down after applying a quantization offset value to coefficient values corresponding to upper high frequencies among all coefficient values of the difference image, and then applying lower low frequency. After applying a quantization offset value to the coefficient values corresponding to, integer quantization may be applied by rounding up or down. In this case, the quantization unit 153 applies a quantization offset of 1 / M to coefficient values corresponding to the lower M low frequencies, and 1 / N to coefficient values corresponding to upper N (N <M) high frequencies. The interval length Δ of the quantization coefficient is adjusted by applying a quantization offset. In operation 640, the quantization unit 153 applies the quantization offset value used in encoding the inter frame to the remaining coefficient values except for the upper M and lower N coefficient values among the total coefficient values.

In addition, in the quantization step, the quantization unit 153 may apply the scanning method described with reference to FIG. 5A or 5B to the difference image. For example, the quantization unit 153 may apply the scanning method in the vertical direction for the 16 × 8 horizontal mode, and apply the scanning method in the horizontal direction for the 8 × 16 vertical mode.

In addition, in the quantization step, when the quantization coefficient value to be currently encoded is 1, the quantization unit 153 considers whether to encode the macroblock to be currently encoded in the difference image in consideration of a quantization coefficient having a previously encoded quantization coefficient value of 0. Can be determined.

In operation 650, the rearrangement unit 155 rearranges coefficient values of the quantized difference image, and the entropy coding unit 157 encodes and compresses the coefficient values of the rearranged difference image. The scanning process described with reference to FIG. 5A or 5B may be performed by the quantization unit 153 or the reordering unit 155.

According to the above-described embodiment of the present invention, the three-dimensional transform performed by the transform unit 151 of the encoding apparatus 150 is an integer DCT, that is, two-dimensional transform, in units of 4 × 4 blocks, and four adjacent four The coefficient values corresponding to the same frequency among the × 4 blocks are collected to perform a one-dimensional transform that performs an integer DCT in units of 4x1 blocks. At this time, since the block size is a small 4x4 block, the difference image error can be reduced, and the spatial redundancy can be reduced by converting twice. Accordingly, the transform unit 151 performs three-dimensional transform in inter prediction with little variation in the difference image error, and the quantization unit 153 quantizes the integer to improve the compression performance.

The encoding apparatus and method for compressing video data by multidimensional transform and integer quantization have been described above. The decoding apparatus and method may also decode a bitstream by integer dequantization and multidimensional inverse transform.

7 is a block diagram illustrating a video data reconstruction apparatus including a decoding apparatus capable of integer inverse quantization and multi-dimensional inverse transformation according to a proposed embodiment.

The decoding apparatus (hereinafter, referred to as a "decoder") 710 illustrated in FIG. 7 reconstructs an image encoded corresponding to the encoder 150 described above.

For example, the decoder 710 may adaptively reconstruct a bitstream to be currently reconstructed using integer quantization information and multidimensional transform information included in the bitstream received from the compression apparatus 100. In addition, the decoder 710 may decode the current image with a decoding size corresponding to the mode information received from the compression apparatus 100.

In this case, the inverse quantization and inverse transform operation in the decoder 710 is substantially the same as the multidimensional transform and integer quantization operation in the encoder 150 of the compression apparatus 100, and the mode information is used in the decoder 710. The decoding operation is substantially the same as the encoding operation of the current image using the encoding mode in the encoding unit 150, and thus redundant description thereof will be omitted.

Referring to FIG. 7, the video data reconstruction apparatus 700 may include a decoder 710, an intra predictor 720, a motion compensator 730, an adder 740, a filter 750, and a memory 760. It may include.

The decoder 710 restores an image by decoding the input bitstream. To this end, the decoder 710 includes an entropy decoder 711, a reordering unit 713, an inverse quantization unit 715, and an inverse transform unit 717.

The entropy decoding unit 711 may entropy decode the input bitstream to extract encoding information such as quantized coefficient values, motion vectors, and mode information of the difference image. The quantized coefficient value of the difference image extracted by entropy decoding is an integer equal to one. In addition, the mode information may include a decoding mode corresponding to the encoding mode determined or changed in units of frames, GOPs, slices, or macroblocks with respect to the current image.

The reordering unit 713 rearranges the quantized coefficient values of the extracted difference image. The rearranged quantization coefficient values have an integer type.

The inverse quantization unit 715 may dequantize the rearranged integer quantization coefficients in macroblock units. Accordingly, the inverse quantization unit 715 inversely quantizes the integer quantization coefficient value and generates coefficient values of the difference image corresponding to the multi-dimensional inverse transform. The inverse quantization unit 715 may perform integer inverse quantization using an inverse quantization table or inverse quantization offset suitable for an Invert Discrete cosine transform (IDCT).

Types of inverse quantization may include frequency domain scalar inverse quantization, spatial domain scalar inverse quantization, frequency domain vector inverse quantization, and spatial domain vector inverse quantization.

When the bitstream is three-dimensionally transformed by the compression apparatus 100, the inverse quantization unit 715 applies integer inverse quantization using an inverse quantization offset to an integer quantization coefficient value, thereby counting the coefficient value of the three-dimensional transformed difference image. Create

The inverse transform unit 717 may obtain the difference image by inversely transforming the coefficient value of the generated difference image. The inverse quantization unit 715 and the inverse transformer 717 may perform different inverse quantization and inverse transformation on each of the macroblocks or subblocks included in the input bitstream.

In particular, the inverse transform unit 717 performs a multidimensional integer IDCT in order to avoid a decimal point operation in the IDCT process, which means that a multidimensional inverse transform (for example, a 3D transform) in the inverse transform unit 717 is based on an integer IDCT. Means.

The intra predictor 720 may predict and decode a macroblock to be decoded using spatially adjacent pixel values in order to encode a block of an entropy decoded image. For example, the intra predictor 325 may generate a prediction block by performing intra prediction.

The motion compensator 730 may generate a prediction block of the current block by using the motion vector extracted by the entropy decoder 711. For example, the motion compensator 730 may generate a prediction block by performing inter prediction.

The adder 740 may reconstruct the current block or the current frame by adding the difference image (ie, the difference block) and the prediction block. The difference image may be input from the inverse transformer 717 to the adder 740, and the prediction block may be input from the intra predictor 720 or the motion compensator 730 to the adder 550.

The reconstructed block or reconstructed frame is stored in the memory 760 through the filter 750 and input to the motion compensator 730 to be used for inter prediction on the next block or the next frame.

8 is a flowchart provided to explain a video decoding method according to an embodiment of the present invention.

In operation 810, the decoding apparatus may entropy-decode the bitstream received from the encoding apparatus to extract encoding information such as quantized coefficient values, motion vectors, and mode information of the difference image. The quantized coefficient values of the extracted difference image have an integer type.

In operation 820, the decoding apparatus rearranges the quantized coefficient values of the extracted difference image.

In operation 830, the decoding apparatus may dequantize the rearranged integer quantization coefficient values in macroblock units. In operation 830, inverse quantization may be performed using an inverse quantization table or an inverse quantization offset suitable for IDCT.

In operation 840, the decoding apparatus may generate a difference image by IDCTing the coefficient value of the generated difference image. In operation 840, the multi-dimensional integer IDCT performing the IDCT process twice may be performed.

The difference image acquired in operation 840 may be added to the prediction block and restored to the current block or the current frame. The prediction block may be a block that is inter predicted by intra prediction or motion compensation.

The method according to the present invention can be implemented in the form of program instructions that can be executed by various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. The medium may be a transmission medium such as an optical or metal line, a wave guide, or the like, including a carrier wave for transmitting a signal designating a program command, a data structure, or the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the claims below but also by the equivalents of the claims.

100: video data compression device
110: intra prediction unit 120: motion estimation unit
130: motion compensation unit 140: first quarter
150: encoder 160: inverse quantization and inverse transform unit
170: second quarter 180: filter
190: frame memory

Claims (10)

A transform unit which multi-dimensionally transforms the difference image generated by removing the predicted image generated by the inter prediction from the original image; And
A quantization unit applying integer quantization to the multi-dimensional transformed difference image;
Encoding apparatus capable of multi-dimensional transformation and quantization comprising a. The method of claim 1,
The quantization unit performs the integer quantization in units of macroblocks. When the quantization coefficient value to be encoded is 1, the quantization unit determines whether to encode the macroblock in consideration of a quantization coefficient having a previously encoded quantization coefficient value of 0. An encoding device capable of multidimensional transformation and quantization. The method of claim 1,
The quantizer is a dead zone by applying the integer quantization by rounding up or down after applying a quantization offset value to coefficient values corresponding to upper high frequency positions among quantization coefficient values of a macroblock of the difference image. A coding apparatus capable of multi-dimensional transformation and quantization that increases a value. The method of claim 1,
The quantizer is a dead zone by applying the integer quantization by rounding up or down after applying a quantization offset value to coefficient values corresponding to lower low frequency positions among quantization coefficient values of a macroblock of the difference image. A coding apparatus capable of multi-dimensional transform and quantization to reduce the noise. The method according to claim 3 or 4,
The quantization unit applies a quantization offset of 1 / M to coefficient values corresponding to lower M low frequencies among the quantization coefficient values, and 1 / N to coefficient values corresponding to upper N (N <M) high frequencies. A coding apparatus capable of multidimensional transformation and quantization by adjusting a dead zone among quantization coefficients by applying a quantization offset of. The method of claim 5,
The quantization unit is capable of multi-dimensional transformation and quantization to apply the quantization offset value used in the encoding of the inter frame to the remaining coefficient values except the upper M and lower N coefficient values of the quantization coefficient values. The method of claim 1,
The quantization unit applies a scanning method in a direction perpendicular to a horizontal direction that is a direction of the horizontal mode for a 16 × 8 horizontal mode, and scans in a direction perpendicular to a vertical direction that is a direction of the vertical mode for an 8 × 16 vertical mode. A coding apparatus capable of multi-dimensional transformation and quantization using a scheme. Multi-dimensional transforming the difference image generated by removing the prediction image generated by the inter prediction from the original image; And
Integer quantization of the multi-dimensional transformed difference image;
A coding method capable of multidimensional transformation and quantization comprising a. An entropy decoding unit for entropy decoding the input bitstream to extract integer quantized coefficient values and motion vectors;
An inverse quantizer for generating coefficient values of a difference image by applying integer inverse quantization to the extracted integer quantized coefficient values; And
An inverse transform unit generating a difference image by multi-dimensional inverse transform of coefficient values of the generated difference image;
Decoding apparatus comprising a. Entropy decoding the input bitstream to extract an integer quantized coefficient value;
Generating coefficient values of a difference image by applying integer inverse quantization to the extracted integer quantized coefficient values; And
Generating a difference image by multi-dimensional inverse transformation of the coefficient values of the generated difference image;
Decryption method comprising a.

Images