CN102970533A - Image cutting system and method - Google Patents

Abstract

An image cutting system and method, the system is used to: cut the largest coding unit LCU into multiple N×N blocks; calculate the angle of each N×N block; count the current coding unit CU block of the LCU The angle of each N×N block; determine the cutting method of the current CU block according to the above statistical results; and if the cutting method of the current CU block is continuous cutting, then cut the current CU block into four sub-CUs of the same size Block, if the cutting mode of the current CU block is to stop cutting, stop cutting the current CU block. The invention can reduce the amount of calculation and speed up the cutting of the four-point tree in the LCU.

Description

Image diced system and method

Technical field

The present invention relates to a kind of image compression system and method, relate in particular to image diced system and method in a kind of image compression.

Background technology

In April, 2010, Joint Collaborative Team on Video Coding(JCT-VC) takes up in compress technique efficient video coding (the HighEfficiency Video Coding of research a new generation, namely H.265/HEVC), its objective is and become follow-on compression standard.Compared to present compress technique H.264, its target is for reducing by 50% bit rate (Bit rate), and its image quality of while and computation complexity have also promoted 3 times.General research direction is at present: promote compression efficiency, improve coding correctness and error recovery capabilities, reduction computing time and computation complexity.

In the coding flow process, image or picture can be cut into maximum coding unit (the Largest Coding Unit of a plurality of fixed sizes, LCU), then can cut into the framework recurrence of four minutes trees the coding unit (Coding Units, CU) of different sizes in each LCU to carry out pixel prediction.Also can judge cutting mode best in this LCU in the process of recurrence cutting, the optimum Cutting mode after recurrence finishes in this LCU will be determined.In this recursive procedure, owing to tested the CU of each different size with the cutting mode of definite the best, the operation time that it is spent and computational complexity thereof are quite high and to be improved.

Summary of the invention

In view of above content, be necessary to provide a kind of image diced system, it can utilize marginal information in the image to prejudge cutting mode in the LCU, and then reduces the number of times of recursive operation in four minutes trees, to reach the purpose that reduces amount of calculation and accelerate tree cutting in four minutes in the LCU.

In view of above content, be necessary to provide a kind of image cutting method, it can utilize marginal information in the image to prejudge cutting mode in the LCU, and then reduces the number of times of recursive operation in four minutes trees, to reach the purpose that reduces amount of calculation and accelerate tree cutting in four minutes in the LCU.

A kind of image diced system is applied to electronic installation, and this system comprises: the first cutting module, and a maximum coding unit LCU for obtaining an image cuts into a plurality of N * N block with this LCU; Computing module is for the angle of calculating each N * N block; Statistical module is for the angle of each N of present encoding unit CU block that adds up this LCU * N block; Judge module, for the cutting mode of determining current C U block according to above-mentioned statistics, described cutting mode comprises the continuation cutting and ends cutting; And the second cutting module, if the cutting mode that is used for current C U block then cuts into current C U block four onesize inferior CU blocks for continuing cutting, if the cutting mode of current C U block, is then ended cutting current C U block for ending cutting.

A kind of image cutting method is applied to electronic installation, and the method comprises: cutting step one, obtain a maximum coding unit LCU in the image, and this LCU is cut into a plurality of N * N block; Calculation procedure is calculated the angle of each N * N block; Statistic procedure, the angle of adding up each N * N block in the present encoding unit CU block of this LCU; Determining step is determined according to above-mentioned statistics that the cutting mode of current C U block, described cutting mode comprise and is continued cutting and end cutting; And cutting step two, if the cutting mode of current C U block then cuts into current C U block four onesize inferior CU blocks for continuing cutting, if the cutting mode of current C U block, is then ended cutting current C U block for ending cutting.

Compared to prior art, described image diced system and method, it can utilize marginal information in the image to prejudge cutting mode in the LCU, and then reduces the number of times of recursive operation in four minutes trees, reduces amount of calculation and accelerates tree cutting in four minutes in the LCU.

Description of drawings

Fig. 1 is the running environment schematic diagram of image diced system of the present invention.

Fig. 2 is the functional block diagram of image diced system.

Fig. 3 is the flow chart of the first embodiment of image cutting method of the present invention.

Fig. 4 is the schematic diagram of determined four minutes tree of the present invention.

Fig. 5 is the schematic diagram that carries out the image cutting according to Si Fenshu shown in Figure 4.

Fig. 6 is the flow chart of the second embodiment of image cutting method of the present invention.

The main element symbol description

Electronic installation	2
		Display device	20
Input equipment	22
		Memory	23
The image diced system	24
		Processor	25
The first cutting module	240
		Computing module	241
Statistical module	242
		Judge module	243
The second cutting module	244
		Prediction module	245

Embodiment

As shown in Figure 1, be the running environment schematic diagram of image diced system of the present invention.This image diced system 24 runs in the electronic installation 2.This electronic installation 2 also comprises display device 20, input equipment 22, memory 23 and the processor 25 that links to each other by data/address bus.Described electronic installation 2 can be computer, mobile phone, PDA(Personal Digital Assistant, personal digital assistant) etc.

Described memory 23 is for data such as the program code of storing described image diced system 24 and images.Described display device 20 is used for showing the data such as described image, the LCDs that this display device 20 can be computer, the touch-screen of mobile phone etc.Described input equipment 22 is used for the various data that the input user arranges, for example, and keyboard, mouse etc.

Described image diced system 24 is used for utilizing the marginal information of image to prejudge the cutting mode of each block in the LCU, when the consistency of each the sub-block direction in judging certain block is higher, stops to continue this block of cutting, and detailed process is following to be described.

In the present embodiment, described image diced system 24 can be divided into one or more modules, described one or more module is stored in the described memory 23 and is configured to and carried out by one or more processors (present embodiment is a processor 25), to finish the present invention.For example, consult shown in Figure 2ly, described image diced system 24 is divided into the first cutting module 240, computing module 241, statistical module 242, judge module 243, the second cutting module 244 and prediction module 245.The alleged module of the present invention is to finish the program segment of a specific function, is more suitable in describing the implementation of software in electronic installation 2 than program.The concrete function of each module is described below with reference to Fig. 3.

As shown in Figure 3, be the flow chart of the first embodiment of image cutting method of the present invention.

Step S10, the first cutting module 240 obtain an image from memory 23 or other electronic equipment, a maximum coding unit in this image (Largest Coding Unit, LCU) is cut into a plurality of N * N block.Be appreciated that, the LCU that comprises a plurality of fixed sizes in image, be 64 * 64 such as size, when this image is encoded, each LCU cuts into the coding unit (CodingUnits of different sizes with the framework recurrence of four minutes trees, CU) block, each coding unit block comprises a plurality of N * N block.

Be appreciated that size is different at the CU block of different levels recurrence generation and the CU block ratio on upper strata.For example, consult Fig. 4 and shown in Figure 5, be cut into four CU block b, g, h and i during this LCU recurrence first time, wherein, first CU block b is continued to cut into again four less inferior CU block c, d, e and f.

Present embodiment describes so that a LCU is cut into example.In the present embodiment, N=8.Take the size 64 * 64 of the most general LCU as example, before the cutting of beginning recurrence, can be cut to first the block of 64 8 * 8 sizes.

Step S11, computing module 241 calculates the angle (such as the angle of 64 8 * 8 blocks) of each N * N block.The method of calculating angle has many kinds, comprises and utilize discrete cosine transform (Discrete Cosine Transform, DCT) or Sobel algorithm etc. that present embodiment describes as an example of the Sobel algorithm example.

Suppose N=8, computing can obtain 64 angles and weight (Weight) thereof to 64 pixels in 241 pairs of 8 * 8 blocks of computing module through Sobel.These 64 angles are mapped to (comprising 34 kinds of patterns) behind the pattern list of infra-frame prediction (Intra prediction), then can count the cumulative rear maximum a certain pattern of weight, and the angle that this pattern corresponds to can represent the angle of this 8 * 8 block, namely represents the marginal information of this 8 * 8 block.Need to prove, as long as can produce an angle in the mode of Representative Region intra-block edge information, all applicable to the present invention.

Step S12, the angle of each N * N block in each coding unit block (hereinafter referred to as " CU block ") of statistical module 242 these LCU of statistics.In the present embodiment, statistical module 242 is according to recursive sequence, adds up successively the angle of each N * N block in each the CU block in each recursive hierarchical of this LCU, for example, adds up 4 CU blocks of ground floor for the first time.

Then, judge module 243 is determined the cutting mode of each CU block in this LCU according to statistics.Described cutting mode comprises the cutting that continues cutting current C U block and end current C U block.Consulting shown in Figure 4ly, is corresponding four minutes trees of the cutting mode schematic diagram of each CU block in determined this LCU of the present invention, and each node represents a CU block.As can be seen from Figure 4, CU block " h " does not need to continue down cutting.

Particularly, if the N that angle is identical or adjacent in the current C U block * N block surpasses preset ratio (such as 50%), then judge module 243 judges that the direction consistency of the N that these current CU blocks include * N block is high, do not need down to cut, namely end the cutting (consulting the CU block " h " among Fig. 4) of current C U block.If the N that angle is identical or adjacent in the current C U block * N block is less than or equal to this preset ratio, then judge module 243 judges that the direction consistency of the N that these current CU blocks include * N block is not high, needs to continue cutting current C U block (consulting the CU block " b " among Fig. 4).In the present embodiment, described angle is adjacent is to determine according to whether intra prediction mode corresponding to angle be adjacent.For example, if the logical order (Logical order) of two intra prediction modes in pattern list is adjacent, judge that then these two corresponding angles of intra prediction mode are adjacent.

For example, suppose that the size of current C U block is 32 * 32, include 16 8 * 8 blocks, namely have 16 directions to represent the marginal information of this CU block.If the consistency of these 16 directions is high, for example identical or adjacent more than half, then can a direction represent this 32 * 32 block, the following recursive program of this 32 * 32 block will be skipped, and carry out the judgement of next 32 * 32 blocks.If these 16 directions are in disorder, for example, identical or adjacent angle is less than half, it is complicated and can not be with a direction representative to represent marginal information in this 32 * 32 block, so recursive program can continue this 32 * 32 block is cut into 4 16 * 16 CU blocks, and each time CU block proceeded to judge last four minutes trees that generate will as shown in Figure 4, be one and set in non-complete four minutes.

In the present embodiment, through step S10-S12, the cutting mode of each CU block pre-determines, and subsequent step will begin each CU block is cut according to predetermined cutting mode.

Step S13, the second cutting module 244 judge according to the cutting mode of each CU block whether current C U block needs to continue cutting.If current C U block need to continue cutting, then execution in step S14-S15; If current C U block does not need to continue cutting, then execution in step S16-S18.

Step S14, the second cutting module 244 cuts into four onesize inferior CU blocks with current C U block.Consult shown in Figure 5, it is the schematic diagram that the Si Fenshu according to Fig. 4 cuts, this LCU comprises four the CU block " b ", " g ", " h ", " i " that wait size, and wherein, CU block " b " continues to cut into four inferior CU block " c ", " d ", " e ", " f " that wait size.

Step S15, the second cutting module 244 choose the upper left degree CU block of current C U block as new current C U block, and then, flow process is returned step S13.Consult shown in Figure 5ly, after CU block " b " was cut into four inferior CU blocks " c " that wait size, " d ", " e ", " f ", the second cutting module 244 was chosen time CU block " c " as new current C U block, continues down to cut.

Step S16, the second cutting module 244 is ended the cutting of current C U block, and 245 couples of current C U of prediction module block carries out pixel prediction.Wherein, described pixel prediction comprises that infra-frame prediction (Intra Prediction) and frame predict (Inter Prediction) outward.In the present embodiment, only carry out infra-frame prediction.

Step S17, the second cutting module 244 judge whether the CU block among this LCU is all predicted complete.If the CU block among this LCU all prediction is complete, then flow process finishes.If also have the CU block to predict among this LCU, then continue execution in step S18.

Step S18, the second cutting module 244 is chosen next CU block as new current C U block according to the zigzag order from this LCU.For example, consult shown in Figure 5ly, according to the zigzag order, if current C U block " e " prediction is complete, then choose the CU block " f " identical with current C U block " e " size as new current C U block.If current C U block " f " prediction is complete, then choose CU block " g " with the sizes such as last layer (being block " b ") of current C U block " f " as new current C U block.

In above-mentioned the first embodiment, statistical module 242 counts the angle of each N * N block in each CU block of this LCU, and judge module 243 pre-determines the cutting mode of each CU block in this LCU according to statistics.Then, the second cutting module 244 is according to the cutting mode of predetermined each CU block, and the CU block that needs are cut cuts.

In a second embodiment, also can adopt while judging the mode of cutting to carry out, i.e. the angle of each N * N block in the current C U block of this LCU of 242 statistics of each statistical module, judge module 243 is determined the cutting mode of this current CU block according to statistics.Then, the cutting mode that the second cutting module 244 is determined according to judge module 243 determines it is to continue cutting current C U block, still ends the cutting of current C U block.Consult shown in Figure 6ly about the flow chart of the second embodiment, the flow process of the flow process of the second embodiment and the first embodiment is basic identical, does not repeat them here.

It should be noted that at last, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, although with reference to preferred embodiment the present invention is had been described in detail, those of ordinary skill in the art is to be understood that, can make amendment or be equal to replacement technical scheme of the present invention, and not break away from the spirit and scope of technical solution of the present invention.

Claims (18)

1. An image cutting system applied to electronic devices, characterized in that the system comprises: The first cutting module is used to obtain a maximum coding unit LCU in an image, and cut the LCU into a plurality of N×N blocks; Calculation module, for calculating the angle of each N×N block; A statistical module, configured to count the angles of each N×N block in the current coding unit CU block of the LCU; A judging module, configured to determine the cutting mode of the current CU block according to the above statistical results, the cutting mode includes continuing cutting and stopping cutting; and The second cutting module is used to cut the current CU block into four sub-CU blocks of the same size if the cutting method of the current CU block is to continue cutting, and if the cutting method of the current CU block is to stop cutting, then Stop cutting the current CU block. 2. The image cutting system according to claim 1, wherein the calculation module uses discrete cosine transform or Sobel algorithm to calculate the angle of each N×N block, wherein N=8. 3. The image cutting system according to claim 2, wherein the calculation module uses the Sobel algorithm to calculate the angle of each N×N block comprising: The 64 pixels in each 8×8 block are subjected to Sobel operation to obtain 64 angles and their weights; After the 64 angles are mapped to the intra prediction mode table, the mode with the largest accumulated weight is calculated, and the angle corresponding to the mode with the largest accumulated weight is selected as the angle of the 8×8 block. 4. The image cutting system according to claim 1, wherein the judging module determines the cutting mode of the current CU block comprising: If the N×N blocks with the same angle or adjacent N×N blocks in the current CU block exceed the preset ratio, it is determined that the cutting method is to stop cutting; If the N×N blocks with the same angle or adjacent to the current CU block are less than or equal to the preset ratio, it is determined that the cutting method is to continue cutting. 5 . The image cutting system according to claim 4 , wherein the angular adjacency is determined according to whether the intra prediction modes corresponding to the angles are adjacent. 6. The image cutting system according to claim 1, wherein the second cutting module is also used for: After the current CU block is divided into four sub-CU blocks of the same size, the upper-left sub-CU block of the current CU block is selected as a new current CU block. 7. The image cutting system according to claim 1, characterized in that the system further comprises: The prediction module is used for performing pixel prediction on the current CU block after cutting the current CU block is stopped. 8. The image cutting system according to claim 7, wherein the pixel prediction includes intra-frame prediction and extra-frame prediction. 9. The image cutting system according to claim 7, wherein the second cutting module is also used for: After judging to stop cutting the current CU block, continue to judge whether all the CU blocks in the LCU have been predicted; If there are still CU blocks in the LCU that have not been predicted, the next CU block is selected from the LCU in zigzag order as the new current CU block. 10. An image cutting method applied to an electronic device, characterized in that the method comprises: Cutting step 1, obtaining a maximum coding unit LCU in an image, and cutting the LCU into a plurality of N×N blocks; Calculation step, calculating the angle of each N×N block; A statistical step, counting the angles of each N×N block in the current coding unit CU block of the LCU; The judging step is to determine the cutting mode of the current CU block according to the above statistical results, and the cutting mode includes continuing cutting and stopping cutting; and Cutting step 2: If the cutting method of the current CU block is to continue cutting, then cut the current CU block into four sub-CU blocks of the same size; if the cutting method of the current CU block is to stop cutting, stop cutting the current CU block. CU block. 11 . The image cutting method according to claim 10 , wherein the calculating step comprises: calculating the angle of each N×N block by using discrete cosine transform or Sobel algorithm, wherein N=8. 12. The image cutting method according to claim 11, wherein the step of calculating the angle of each N×N block using the Sobel algorithm comprises: The 64 pixels in each 8×8 block are subjected to Sobel operation to obtain 64 angles and their weights; After the 64 angles are mapped to the intra prediction mode table, the mode with the largest accumulated weight is calculated, and the angle corresponding to the mode with the largest accumulated weight is selected as the angle of the 8×8 block. 13. The image cutting method according to claim 10, wherein the judging step comprises: If the N×N blocks with the same angle or adjacent N×N blocks in the current CU block exceed the preset ratio, it is determined that the cutting method is to stop cutting; If the N×N blocks with the same angle or adjacent to the current CU block are less than or equal to the preset ratio, it is determined that the cutting method is to continue cutting. 14 . The image segmentation method according to claim 13 , wherein the angular adjacency is determined according to whether the intra prediction modes corresponding to the angles are adjacent. 15. The image cutting method according to claim 10, further comprising: After the current CU block is divided into four sub-CU blocks of the same size, the upper-left sub-CU block of the current CU block is selected as a new current CU block. 16. The image cutting method according to claim 10, further comprising: The prediction step is to perform pixel prediction on the current CU block after cutting the current CU block is stopped. 17. The image segmentation method according to claim 16, wherein the pixel prediction includes intra-frame prediction and extra-frame prediction. 18. The image cutting method according to claim 16, further comprising: After judging to stop cutting the current CU block, continue to judge whether all the CU blocks in the LCU have been predicted; If there are still CU blocks in the LCU that have not been predicted, the next CU block is selected from the LCU in zigzag order as the new current CU block.

Images