CN1312933C - A video image compression coding method based on dendritic structure - Google Patents

Abstract

The present invention relates to an improved method for a video image compression code, which has high performance and is based on a tree structure. An SPIHT is a simple algorithm for compression codes with a high efficiency, and however, the algorithm has the defects of large requirement of storage space and difficult hardware implementation, which always impedes the further development and the application of the algorithm. A literature [2] provides an SPIHT encoding algorithm based on a tree structure. On the basis, the paper introduces a threshold estimation cut' technology to improve the SPIHT encoding algorithm, which further reduces memory consumption needed when the SPIHT encoding algorithm is realized and enhances the coding speed so that the SPIHT algorithm is easier to realize. An experiment verifies that the method can more effectively reduce the storage space needed for the realization of the algorithm and reduce time consumption under the precondition that a compression effect is not affected. The present invention solves the problem of difficult SPIHT hardware transplant.

Description

A kind of video image compression coding method based on tree

Technical field

The invention belongs to the video image compression technology field, be specifically related to a kind of high performance, improve one's methods based on the video image compression coding of tree.

Background technology

SPIHT (set partitioning in hierarchical tree) ^[1]Be a kind of simply, ZT image compression encoding algorithm efficiently, it is preserved the pixel coordinate by middle chained list and simplifies scanning process to wavelet coefficient, thereby has improved compression efficiency greatly and realized simplicity.But the middle chained list of spiht algorithm has taken a large amount of memory spaces, is unfavorable for the hardware realization, hampers spiht algorithm further development and application always.Therefore, how under the prerequisite that does not reduce compression effectiveness, shorten chained list length, the memory space that reduces internal memory has become the focus of recent research.

Document [3] has proposed " STTP-SPIHT (Spatio-Temporal Tree Preserving 3-D SPIHT) algorithm ", has at first mentioned the notion that D S PIHT is divided according to space and time-domain.Document [2] has been considered the hardware transplanting problem of SPIHT, " (SPIHT) video image compressing method is divided in the hierarchical tree set based on tree " proposed, it is the N two field picture to be divided into M tree handle respectively, this handles the N two field picture in original spiht algorithm and compares, and can save a lot of spaces.But, the shortcoming of this method also clearly, it can not the accurate algorithm of image scale requires block code stream in real time according to different compressions like that, and after must waiting until that all trees all dispose, could pass through corresponding code stream allocation algorithm, obtain every actual code stream length that transmits of tree.Therefore, in advance and when not knowing the actual code stream length of every tree, for the correctness that guarantees to encode, must all carry out lossless compress to every tree, the code stream of corresponding every the tree that obtains can not be exported immediately, and must leave in the memory, after treating that all trees all dispose, synthesize the output of target code stream again.From the test result of document [2], we are not difficult to find: do not contact directly between chained list length and the compression ratio, chained list length can't reduce along with the increase of compression ratio.This is because even under the condition of high compression ratio, also must carry out lossless coding to all trees.There is very big redundancy in this method in the process of every tree being carried out lossless compress.Promptly this algorithm does not still solve the excessive problem of chained list length, and the internal memory use amount after the reduction is still sizable, thereby is only limited to and is applied on the big hardware platform of some internal memories, and the hardware that can not really solve SPIHT is transplanted problem.

The present invention has proposed " interceptive value prediction " and has improved one's methods on the basis of document [2], it is minimum that the internal memory use amount is dropped to, and further improves arithmetic speed simultaneously, makes spiht algorithm go for general hardware platform.

List of references (References)

[1]A.Said and W.A.Pearlman，“ANew Fast and Efficient Image Codec Based on SetPartitioning in Hierarchical Trees，”IEEE Trans.Circ.and Syst.for Video Technology，Vol.6，pp.243-250，June 1996

[2] patent: the video image compressing method application number is divided in the hierarchical tree set based on tree: 200410017558.7 inventors: magnificent Yun, Hu Bo, Xu Sheng, Gao Jia

[3]S.Cho and W.A.Pearlman.Error resilient compression andtransmission of scalable video.Applications of Digital ImageProcessing XXIII，Proceedings SPIE，4115，2000.

Summary of the invention

The objective of the invention is to propose a kind of video image compressing method based on tree, this method can reduce the use amount of internal memory, and further improve arithmetic speed as far as possible guaranteeing farthest to reduce redundant information under the prerequisite that compression effectiveness does not descend.

The SPIHT video image compressing method that the present invention proposes based on tree, concrete steps are as follows: show according to " adding up minimum threshold value ", with " adding up minimum threshold value " under a certain compression ratio as interceptive value, Y component (or U component or V component) to the preceding N frame of video image carries out lossy compression method, obtain the actual interceptive value under this compression ratio, obtain interceptive value under other compression ratios, other frames and other picture contents according to " interceptive value method of estimation " again, and carry out compressed encoding respectively.

In the said method, described " interceptive value method of estimation " is as follows:

(1) with the interceptive value under a certain picture content of N continuous frame under any compression ratio, as the interceptive value under other picture contents of other N continuous frames under other compression ratio;

(2) interceptive value under the known a certain compression ratio is estimated interceptive value under other compression ratios according to the proportionate relationship shown in the table 1;

(3) under same compression ratio, with the interceptive value of certain N continuous two field picture interceptive value as other N continuous two field pictures;

(4) with the interceptive value of a certain picture content under the same compression ratio interceptive value as other picture contents.

The inventive method proposes to be based on following statistics.

1, there is minimum in the interceptive value of video image

We add up the interceptive value of frame video images up to a hundred, have drawn under different compression ratios, are applicable to " adding up minimum threshold value " table (seeing Table 1) of general video image.The interceptive value of general pattern under certain compression ratio is necessarily more than or equal to " adding up minimum threshold value " of correspondence.Use this table, interceptive value is made as " adding up minimum threshold value ", then can not influence fully under the condition of compression effectiveness, the lossless compress of every tree is converted into lossy compression method, thereby save the internal memory use amount and accelerate arithmetic speed.

Compression ratio	Add up minimum threshold value
		50	4
100	8

200	16
		400	32

Table 1 " is added up minimum threshold value " and is shown

2, be the substantial linear proportionate relationship between interceptive value and the corresponding compression multiple.

We test tens kinds of different images, find that identical video image is under different compression multiples (compression ratio is less than 400) require, behind " (SPIHT) divided in the hierarchical tree set based on tree " compressed encoding, roughly present the linear ratio relation between resulting interceptive value and the corresponding compression multiple.For example: generally speaking, suppose that compression ratio is that 100 o'clock interceptive value is 8, then compression ratio is that 200 o'clock interceptive value just is 16, is 32 and compression ratio is 400 o'clock a interceptive value.Therefore, as long as obtain actual interceptive value under some compression ratios, just can estimate the interceptive value under other compression ratios under a proportional relationship.In actual applications, can obtain compression ratio earlier and be 400 o'clock interceptive value, estimate the interceptive value under the desired compression ratio again, can reduce the internal memory use amount greatly like this.

3, identical compression multiple requires down, and the interceptive value between the picture frame group is roughly the same.

We test a series of continuous video frame image groups, discovery is under identical compression multiple requires, after the N continuous frame video image passed through " (SPIHT) divided in the hierarchical tree set based on tree " compressed encoding, resulting interceptive value was roughly the same through the interceptive value that obtains behind the identical compressed encoding with other continuous N frame video image.Therefore, if the actual interceptive value of N two field picture before obtaining, just can be with its interceptive value as other N continuous frame video images under the same compression ratio.

4, identical compression multiple requires down, and the interceptive value between video image YUV component is roughly the same.

We found that at identical compression multiple to require down that to the compressed encoding that the YUV component of video image carries out " based on the hierarchical tree set division (SPIHT) of tree " respectively YUV component interceptive value separately is roughly the same.Therefore, as long as obtain the actual interceptive value of Y (or U, V) component, just can be with its interceptive value as other two components under the same compression ratio.

Therefore, according to The above results, the present invention proposes based on interceptive value method of estimation in the SPIHT video image compression algorithm of tree.It utilizes some correlation properties between video image, the internal memory use amount when farthest having reduced the SPIHT computing, and further improved arithmetic speed.

Embodiment

1, for one group of new video image, earlier Y (or U, V) component of preceding N two field picture being carried out compression ratio is 400 the SPIHT coding based on tree, and wherein the interceptive value of every tree is set according to " adding up minimum threshold value " in the table 1.Coding finishes, and obtains compression ratio and be 400 o'clock actual interceptive value Thd ₄₀₀

2, according to " interceptive value method of estimation ", extrapolate $\mathrm{Th}{d}_{200}=\frac{1}{2}\mathrm{Th}{d}_{400}$ $\mathrm{Th}{d}_{100}=\frac{1}{4}\mathrm{Th}{d}_{400}$ $\mathrm{Th}{d}_{50}=$ $\frac{1}{8}\mathrm{Th}{d}_{400}$ Use the interceptive value that these estimate, carry out the SPIHT coding under each compression ratio respectively based on tree.

3, for other the continuous N two field picture under the identical compression ratio and the video image of other two components, also, estimate corresponding interceptive value according to " interceptive value estimation " method, carry out SPIHT coding based on tree.

Simulation result:

Concrete simulated conditions is as follows:

Y, the U of MissAmerican video image group 1-48 two field picture, V data (4: 1: 1 sample formats), every two field picture size is 352 * 288.Carry out three grades of 3 d-dem wavelet transformations, again low-frequency frame is carried out the two-stage two-dimensional discrete wavelet conversion, wavelet basis is selected Daubechies 9/7 biorthogonal wavelet (line direction and column direction) and Haar small echo (time orientation) for use.Have 99 trees.

Experimental result is as follows:

Compression multiple	Index	Before the improvement	After the improvement
				400	LIP chained list length (element number)	4488	1203
LSP chained list length (element number)	6416	710
			LIS chained list length (element number)		536	288
The PSNR effect	36.3623	36.2422
			Internal memory saving rate		80.8％
200	LIP chained list length (element number)	4433					2129
			LSP chained list length (element number)	6459	1529
	LIS chained list length (element number)	536				408
			The PSNR effect	39.0733	39.0267
	Internal memory saving rate	64.4％
				100	LIP chained list length (element number)	4443	3521
LSP chained list length (element number)	6448	3032
			LIS chained list length (element number)		536	448
The PSNR effect	40.4405	40.4044
			Internal memory saving rate		62.4％
50	LIP chained list length (element number)	4390					3521
			LSP chained list length (element number)	6462	3032
	LIS chained list length (element number)	536				465
			The PSNR effect	41.3303	40.3002
	Internal memory saving rate	38.4％

Required maximum length when the chained list length of being added up is computing.

According to result of upper experiment, we can clearly find, the SPIHT image encoding algorithm based on tree before high performance SPIHT image encoding based on tree proposed by the invention is improved algorithm and improved is compared, on average can save nearly 60% internal memory use amount, and the PSNR index has only reduction slightly.

Claims (1)

1, a kind of video image compression coding method based on tree, it is characterized in that showing according to " adding up minimum threshold value ", with " adding up minimum threshold value " under a certain compression ratio as interceptive value, the Y component of N frame before the video image or U, V component are carried out lossy compression method, obtain the actual interceptive value under this compression ratio, predict according to " interceptive value method of estimation " again and interceptive value under other compression ratios, other frames and other picture contents carry out compressed encoding respectively; Described " adding up minimum threshold value " table is as follows:

Compression ratio Add up minimum threshold value 50 4 100 8 200 16 400 32

Described " adding up minimum threshold value " is the interceptive value minimum value of adding up the image lossy compression method coding that obtains, and described interceptive value is the threshold value when reaching the target code stream, and described " interceptive value method of estimation " is as follows:

(1) with the interceptive value under a certain picture content of N continuous frame under any compression ratio, as the interceptive value under other picture contents of other N continuous frames under other compression ratio;

(2) interceptive value under the known a certain compression ratio is estimated interceptive value under other compression ratios according to the proportionate relationship shown in the table 1;

(3) under same compression ratio, with the interceptive value of certain N continuous two field picture interceptive value as other N continuous two field pictures;

(4) with the interceptive value of a certain picture content under the same compression ratio interceptive value as other picture contents.