KR101311527B1 - Video processing apparatus and video processing method for video coding - Google Patents

Abstract

An image processing apparatus for performing transformation and quantization processing, comprising: a plurality of transform blocks, the plurality of transform blocks are configured in parallel, and each transform block includes a transform unit for converting input data, and a plurality of transform blocks, respectively. A plurality of quantization blocks, each quantization block including a quantization unit for quantizing data transmitted from a corresponding transform block, and among the plurality of transform blocks and the plurality of quantization blocks, a size of a transform unit; As a result, some blocks are activated and others are deactivated.

Description

Image processing device and image processing method {VIDEO PROCESSING APPARATUS AND VIDEO PROCESSING METHOD FOR VIDEO CODING}

The present invention relates to an image processing apparatus for high efficiency video coding.

Recent multimedia services require high definition (HD) or ultra high definition (UHD) video. So, the recent standardization bodies, Moving Picture Experts Group (MPEG) and Video Coding Experts Group (VCEG), have started the Joint Collaborative Team on Video Coding (JCT-VC) standardization meeting, which is the next generation of video compression standard, High Efficiency. Video Coding (HEVC).

In HEVC, a transform unit (TU) that performs basic transform and quantization is sized according to the size of a coding unit (CU) or a prediction unit (PU). The maximum size of the transform unit (TU) is 32x32, the minimum size is 4x4, and is divided according to the value of N determined by the coding unit CU and the prediction unit PU. In this case, there may be various sizes of transform units (TUs) available according to various coding unit (CU) sizes. Therefore, conventional transducers have to be designed with separate hardware for each conversion unit (TU) size. In addition, in the processing of the conversion unit (TU), it is necessary to design and improve the structure based on the multi-core for the processing of the ultra high resolution image.

SUMMARY OF THE INVENTION The present invention has been made in an effort to provide an image processing apparatus and an image processing method for adaptively transforming transform units having various sizes based on a plurality of transform blocks and a plurality of quantization blocks configured in parallel.

An image processing apparatus for performing transformation and quantization processing according to an embodiment of the present invention, comprising: a plurality of transform blocks, the plurality of transform blocks are configured in parallel, and each transform block is a transform for transforming input data. And a plurality of quantization blocks corresponding to each of the plurality of transform blocks, each quantization block including a quantization unit for quantizing data transferred from a corresponding transform block, and the plurality of transform blocks and the plurality of quantization blocks. Among the blocks, some blocks are activated and others are deactivated according to the size of the transform unit.

The image processing apparatus may further include a controller configured to activate a predetermined number of the plurality of transform blocks and the plurality of quantization blocks based on the size of a transform unit processed by the transform unit.

The image processing apparatus may include a dual memory, and may further include a memory unit for outputting data in parallel and transferring the data to the plurality of conversion blocks.

The transform unit may use a transform basis vector whose value is set to predict the remaining information required for the conversion process based on some information.

The converter may use a partial butterfly operator.

A method of performing a transform and quantization process by an image processing apparatus according to another embodiment of the present invention, the first active block group to be used for transform processing based on the size of a transform unit among a plurality of transform blocks configured in parallel Determining a second activation block group corresponding to the first activation block among a plurality of quantization blocks respectively corresponding to the plurality of transform blocks, wherein the first activation block group and the second activation block are determined. Activating the group, inputting and converting data in parallel to the first activation block group, and inputting and quantizing the output data of the first activation block group to the corresponding second activation block group, respectively. Include.

The transform block may transform using a transform basis vector whose value is set to predict the remaining information necessary for the transform process based on some information.

The transform block may use a partial butterfly operator.

According to an embodiment of the present invention, even if a conversion unit of various sizes is input, the conversion block may be selectively activated to perform conversion processing, and the input data may be processed in parallel to improve data processing performance. In addition, according to an embodiment of the present invention, the lookup table size may be reduced based on the repeatability of the transform basis vector.

1 is a block diagram of an image processing apparatus according to an embodiment of the present invention.
2 is a detailed block diagram of an image processing apparatus according to an embodiment of the present invention.
3 is a detailed block diagram of an image processing apparatus according to another exemplary embodiment of the present invention.
4 is a diagram illustrating a transform transform basis vector according to an embodiment of the present invention.
5 is a view for explaining a butterfly calculation unit according to an embodiment of the present invention.
6 is a block diagram related to quantization processing according to an embodiment of the present invention.
7 is a timing diagram related to transform and quantization according to an embodiment of the present invention.
8 is a flowchart of an image processing method according to an embodiment of the present invention.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like parts throughout the specification.

Throughout the specification, when a part is said to "include" a certain component, it means that it can further include other components, without excluding other components unless specifically stated otherwise.

Now, an image processing apparatus for high efficiency video coding according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings. MUX shown in the drawings is a general device for separating and integrating signals.

1 is a block diagram of an image processing apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the image processing apparatus 100 performs a transform and quantization process for High Efficiency Video Coding (HEVC). The image processing apparatus 100 includes an input unit 200, a memory unit 300, a converter 400, a quantization unit 500, an output unit 600, and a controller 700.

The input unit 200 receives data for transform and quantization processing.

The memory unit 300 classifies and stores the data received from the input unit 200 in a plurality of memory banks. In this case, the memory unit 300 may have a dual memory structure.

The transform unit 400 transforms the data received from the memory unit 300. In this case, the transform unit 400 includes a plurality of transform blocks in parallel. Each of the conversion blocks receives data from the memory unit 300 and simultaneously converts the data. The HEVC transformation structure is shown in Table 1. If the size of the coding unit (CU) is 64X64, the size of the coding unit (CU) possible is 64X64, 32X32, 16X16, 8X8, and the available conversion unit (TU) varies from 4X4 to up to 32X32. The conversion is applied in three stages in high efficiency mode, and in two stages in low complexity mode.

HEVC High Efficiency HEVC Low Complexity Prediction uints (N = 4, 8, 16, 32, shapes: 2Nx2N, NxN for smallest) Transforms can cross prediction unit boundaries for intra; not for inter TU tree Structure (Max 3 Levels) TU Tree Structure (Max 2 Levels) Transform Block size of 4x4 to 32x32 Samples (always Square)

The converter 400 converts signals in a spatial domain, which are dispersed in a compaction of energy, into a frequency domain and uses main components. The transform unit 400 may use various transformation methods. For example, a grid DCT based transform, a grid DST based transform, and a non-grid quadtree transform may be used. tree transform, NSQP). In HEVC, from 8X8 up to 32X32 square blocks are used. In non-lattice quadtree transformations, 4X16, 16X4, 8X32, 32X8 sizes may be used.

The quantization unit 500 quantizes the data received from the conversion unit 400. In this case, the quantization unit 500 includes a plurality of quantization blocks in parallel. Each of the quantization blocks receives data from a corresponding transform block of the transform unit 400 and simultaneously performs quantization processing.

The quantization unit 500 basically performs quantization by dividing the transform coefficients in the transform unit (TU) into tables in order to reduce the magnitude of the transform coefficient obtained by converting the residual error values predicted. Quantization of HEVC is applied to each of luminance and chrominance components by the size of a transform unit (TU). The quantization and inverse quantization tables of HEVC are composed of the same value and vary according to the quantization parameter (QP) value. The quantization unit 500 changes the QP value that matches the order of the given QP array based on the QP value, the current transform unit (TU) size, and the chroma component value. The quantization unit 500 uses the QP value as it is in the case of luma.

The output unit 600 outputs the quantized data received from the quantization unit 500.

The controller 700 activates or deactivates the transform blocks of the converter 400 based on the size of the data input to the input unit 200. Similarly, the control unit 700 activates or deactivates the quantization block corresponding to the transform block of the transform unit 400. For example, when the control unit 700 processes a 4 × 4 transform unit (TU), the transform blocks and the quantization blocks required for the 4 × 4 size processing are activated, and the remaining transform blocks and the quantization blocks are deactivated. Through this, the control unit 700 controls the conversion unit (TU) of various sizes to be converted by the conversion unit 400.

The controller 700 generates a signal related to the QP change and transmits the signal to the input unit 200. The controller 700 may transmit a QP change enable signal to the input unit 200 based on the states of the converter 400 and the quantizer 500. The control unit 700 may change the QP value based on the size of the conversion unit (TU). The input unit 200 receiving the change may change the QP value and transmit the QP value to the quantization unit 500.

2 is a detailed block diagram of an image processing apparatus according to an embodiment of the present invention.

Referring to FIG. 2, the converter 400 and the quantizer 500 process data in parallel. The transform unit 400 includes a plurality of transform blocks 40, and the plurality of transform blocks 40 are configured in parallel. Here, each of the conversion blocks 40 receives data from the memory unit 300 and simultaneously converts the data. The conversion unit 400 may include, for example, 33 conversion blocks 40 to process a conversion unit (TU) having a maximum size of 32 × 32.

When the discrete cosine transform (DCT) is a discrete cosine transform (DCT), the transform unit 400 is a variable structure of a tree structure, and thus, the transform unit 400 is designed to select a mode based on an index. The converter 400 is hardware designed in a sub-matrix structure. In this case, the conversion unit 400 supports full factorization, designs a partial butterfly structure, reduces hardware complexity, and efficiently supports matrix multiplication.

Large transform units (TUs) are often used in images with almost no motion, and large transform units (TUs) are used. On the contrary, many transform units (TUs) are selected for images with many movements or complex textures. . In the HEVC codec, it is difficult to predict a lot of motion and a small amount, so the 32X32, which is a large conversion unit (TU), uses the basic DCT, and the smaller 16X16, 8X8, and 4X4 are 32X32 to satisfy all sizes of the conversion unit (TU). Because it can be processed in DCT block, it can be shared without using separate block. In this case, in the case of the small conversion unit TU, the number of images to be processed increases even when the video of the same size is processed, so that the plurality of conversion blocks 40 are used for parallel processing.

The quantization unit 500 includes a plurality of quantization blocks 50, and the plurality of quantization blocks 50 are configured in parallel. Here, each of the quantization blocks 50 receives data from the corresponding transform block 40 and quantizes them at the same time. In this case, the quantization unit 500 may perform quantization by using the lookup table of the quantization lookup table 800. The quantization lookup table 800 may include a quantization lookup table updater 80 and quantization lookup tables 81 and 82.

FIG. 3 is a detailed block diagram of an image processing apparatus according to another embodiment of the present invention, FIG. 4 is a diagram illustrating a transform transform basis vector according to an embodiment of the present invention, and FIG. 5 is an embodiment of the present invention. It is a figure explaining the butterfly calculation part which concerns on.

Referring to FIG. 3, the memory unit 300 includes a plurality of memory banks 30. The memory unit 300 divides the data received from the input unit 200 into a predetermined size and stores the data in the plurality of memory banks 30 in order.

The transform block 40 includes a transform lookup table 41, transform operators 42 and 43, and a butterfly operator 44. The transform block 40 performs a transform process based on the transform lookup table 41 based on the transform basis vector. In addition, the conversion block 40 reduces the complexity of the operation by using the butterfly operation unit 44.

Referring to FIG. 4, the transform basis vector is composed of a pattern for predicting the remaining information using some information using repeatability and regularity. For example, the transform basis vector may be configured with a value that is symmetrical about the center point or origin symmetry. Thus, knowing the value of any one of the transform basis vectors, e. Therefore, the conversion block 40 can be reduced to 1/2 or 1/4 of the lookup table used so far. As a result, the transform block 40 may convert the data to a value of about 1/4 of the required transform lookup table size.

Referring to FIG. 5, the butterfly operation unit 44 uses a partial butterfly structure using simple addition and shift operations on input residual data during parallel processing of hardware. Therefore, the butterfly operator 44 can reduce the complexity of the operator.

Referring back to FIG. 3, the quantization block 50 receives the transformed data from the corresponding transform block 40. In this case, the quantization block 50 receives a QP value and transform results.

6 is a block diagram related to quantization processing according to an embodiment of the present invention.

Referring to FIG. 6, the quantization block 50 receives data from a corresponding transform block 40 and simultaneously quantizes the data. In this case, the quantization block 50 performs quantization by using the lookup table of the quantization lookup table 800.

The quantization block 50 uses the quantization lookup table unit 800 calculated in advance. That is, since the quantization block 50 requires a value such as a position factor (PF), a multiplication factor (MF), qbits, f, etc., a large amount of computational values, for example, qbits and f values, are calculated in advance and quantized lookup. Save it to a table.

When the quantization block 50 uses the quantization lookup table 800, a modular operation is required to calculate the quotient and the remainder of the QP. Since the modular operation can be implemented using only a comparator and an adder, the complexity of the quantization unit 500 can be reduced.

7 is a timing diagram related to transform and quantization according to an embodiment of the present invention.

Referring to FIG. 7, the image processing apparatus 100 operates according to a timing diagram for input data, output data, and an interface with an upper module in a transform and quantization process. For example, it is assumed that a 32X32 size conversion unit (TU) is input to the input unit 200 as input data together with an input data enable signal.

The controller 700 inputs a QP value along with data input, and outputs a quantization LUT ready signal when a quantization lookup table for quantization is prepared according to the QP value. The transform unit 400 converts the input data, and the quantization unit 500 performs quantization based on the QP value. The converter 400 and the quantizer 500 process data with a difference of 32 clocks. The output unit 600 outputs the data processed by the converter 400 and the quantization unit 500 together with an output data enable signal.

If the QP change enable signal is high, the converter 400 and the quantization unit 500 can process the next data, and if the QP change enable signal is low, Since the conversion unit 400 and the quantization unit 500 are still processing data, the input data cannot be received.

When the input unit 200 receives input data in parallel, 32 clocks are required. When the input unit 200 receives serially, the input unit 200 takes 32X32 clocks. Therefore, it is possible to design flexibly according to the application of the image processing apparatus 100.

8 is a flowchart of an image processing method according to an embodiment of the present invention.

Referring to FIG. 8, the image processing apparatus 100 determines a first active block group to be used for the transform process based on the size of the transform unit TU among the plurality of transform blocks 40 configured in parallel (S810). For example, when 4 × 4 size data is input, the image processing apparatus 100 may determine four transform blocks 40 as active block groups. Similarly, when 8 × 8 size data is input, the image processing apparatus 100 may determine eight transform blocks 40 as active block groups. Herein, the transform block 40 is configured in parallel as described with reference to FIGS. 2 to 5, and the size of the lookup table using a transform basis vector whose value is set to predict the remaining information required for the conversion process based on some information. Reduce The conversion block 40 reduces the complexity of the operation by using the partial butterfly structure.

The image processing apparatus 100 determines a second activation block group corresponding to the first activation block among the plurality of quantization blocks 50 respectively corresponding to the plurality of transformation blocks 40 (S820).

The image processing apparatus 100 activates the first activation block group and the second activation block group (S830).

The image processing apparatus 100 inputs and converts data in parallel to the first active block group (S840).

The image processing apparatus 100 inputs and quantizes the output data of the first activation block group to the corresponding second activation block group (S850).

As described above, according to the exemplary embodiment of the present invention, even when a conversion unit having various sizes is input, the conversion block can be selectively activated to perform conversion processing, and the input data can be processed in parallel to improve data processing performance. In addition, according to an embodiment of the present invention, the lookup table size may be reduced based on the repeatability of the transform basis vector.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, It belongs to the scope of right.

Claims (8)

An image processing apparatus for performing transformation and quantization processing,
A plurality of transform blocks, wherein the plurality of transform blocks are configured in parallel, and each transform block is a transform unit for converting input data, and
A quantization block includes a plurality of quantization blocks corresponding to each of the plurality of transform blocks, and each quantization block quantizes data transferred from the corresponding transform block.
Lt; / RTI >
And among the plurality of transform blocks and the plurality of quantization blocks, some blocks are activated and others are deactivated according to the size of a transform unit. In claim 1,
A controller for activating a predetermined number of the plurality of transform blocks and the plurality of quantization blocks based on the size of a transform unit processed by the transform unit
Image processing apparatus further comprising. In claim 1,
Dual memory, memory unit for outputting data in parallel to the plurality of conversion blocks
Image processing apparatus further comprising. In claim 1,
The conversion unit
And a transform basis vector configured with a pattern in which some information is repeated and whose value is set to predict the remaining information necessary for the conversion process based on the partial information. In claim 1,
And the conversion unit uses a partial butterfly calculator. As the image processing apparatus performs the transform and quantization processing,
Determining a first active block group to be used for the transform process based on the size of a transform unit among a plurality of transform blocks configured in parallel,
Determining a second activation block group corresponding to the first activation block from among a plurality of quantization blocks respectively corresponding to the plurality of transform blocks;
Activating the first activation block group and the second activation block group;
Inputting and converting data into the first active block group in parallel; and
Inputting and quantizing the output data of the first activation block group to the corresponding second activation block group, respectively.
Image processing method comprising a. The method of claim 6,
The transform block is
And converting using a transform basis vector configured with a pattern in which some information is repeated and having a value set to predict the remaining information necessary for the conversion process based on the partial information. The method of claim 6,
The transform block is
Image processing method using partial butterfly calculator.

Images