CN104618718A - Space-time multi-prediction mode based lossless compression method and system - Google Patents

Abstract

The invention discloses a space-time multi-prediction mode based lossless compression method. The space-time multi-prediction mode based lossless compression method comprises step 1, obtaining a sequence image f through a satellite-borne imaging system; step 2, dividing every frame of image fk obtained in the step 1 into sub-blocks fk,i which are mutually non-overlapped, wherein the formulas are as follows, the size of every sub-block is M*N, SumI is the total number of the sub-blocks, and M and N are preset values; step 3, serving the 3j+1<th> frame of image in the step 2 as a reference frame image and performing sub-block JPEG-LS coding, wherein the formula is as follows; step 4, enabling the 3j+2<th> frame of image and the 3j+3<th> frame of image to the 3j+1<th> frame of image for space-time multi-prediction inter-frame lossless coding. The invention also provides a space-time multi-prediction mode based lossless compression system. According to the space-time multi-prediction mode based lossless compression method, the space and time correlation of the image is integrated to improve the prediction mode, the image is divided into the plurality of sub-blocks, the optimal prediction modes of the different sub-blocks are self-adaptively selected for prediction, and accordingly the self-adaption to different characteristics of areas of the image can be achieved through a predictor and accordingly the compression effect on the sequence image is good.

Description

A kind of lossless compression method based on the many predictive modes of space-time and system

Technical field

The invention belongs to digital image processing techniques field, be specifically related to a kind of lossless compression method based on the many predictive modes of space-time and system.

Background technology

Along with the raising of spaceborne imaging load categories and resolution, the image data amount that satellite obtains in effective observation time section is increasing.By restrictions such as storage resources, down biography bandwidth ability on grounded receiving station geographical distribution, star, the view data of magnanimity causes great pressure to satellite data management, and carrying out spaceborne image compression is the inevitable choice solving this problem.Except obtaining the cost height of satellite image, view data itself is also very important, and therefore, general spaceborne compressibility often adopts JPEG-LS lossless compressiong.

JPEG-LS is the ISO/ITU standard for continuous-tone image Lossless Compression, is effective unification rest image being achieved to low complex degree and high compression ratio, but in spaceborne imaging transmission technology, movable image is the main object that satellite is paid close attention to.The temporal image sequence that satellite image is made up of the consecutive image with frame period being interval the time.In sequence image, image is in time than spatially usually having larger correlation, and the difference of adjacent two two field pictures is only the movement of target sometimes.

The simplest computational methods of image frame-to-frame correlation are gray values that direct present frame deducts previous frame image, and namely select the predicted value of pixel value as frame under process of previous frame image correspondence position, the method is encoded also referred to as inter-frame difference.If when image motion amount is little or spatial detail is little, inter-frame difference coding code efficiency very high, but when image spatial detail enrich or when there is noise, inter-frame difference code efficiency can reduce.When there is a large amount of body in motion in image, simple difference prediction can not receive good effect, and as shown in Figure 1, two width images are target travel, and direct differential not only can not make the error range of prediction diminish, and may increase the dynamic range of data.So the inter-frame prediction method with motion compensation is suggested, as shown in Figure 2, the method makes predicated error relatively more accurate by first carrying out estimation to its flow process, thus the error of prediction also can reduce, and can obtain better data compression ratio.Consider the different qualities of image zones of different, and the difference of target area and background area, the inter-frame prediction method with motion compensation is divided into image the sub-block of the formed objects of many non-overlapping copies, different sub-block can an adaptively selected optimum prediction module, as shown in Figure 3, adopt this template to predict, make the accumulative predicated error absolute value in sub-block minimum.If when image space details seldom or target travel amount is less, the code efficiency with the inter prediction encoding of motion compensation is very high, but when the spatial detail of image is enriched, sometimes the frame-to-frame correlation of sub-block is also poorer than spatial coherence, and the inter prediction encoding efficiency with motion compensation can reduce.

In sum, need to study new digital image processing techniques, improve the compression ratio of sequence image, reduce the pressure that spaceborne imaging system stores and transmits data.

Summary of the invention

The object of the present invention is to provide a kind of lossless compression method based on the many predictive modes of space-time, the method compensate for conventional compression method and just utilizes in sequence chart picture frame or the blindness of frame-to-frame correlation, make full use of the correlation of sequence image on room and time, thus effectively can improve image coding efficiency.

To achieve these goals, according to a first aspect of the present invention, provide a kind of lossless compression method based on the many predictive modes of space-time, its step comprises:

(1) spaceborne imaging system is utilized to obtain sequence image f;

(2) each two field picture f that step (1) obtains _k, k=1,2 ... be divided into non-overlapping copies and size is the sub-block f of M × N _k,i, i=1,2 ..., SumI, SumI are sub-block sum, and M, N are preset value;

(3) 3j+1, j=0 in step (2) is made, 1 ... two field picture is reference frame image, carries out piecemeal JPEG-LS coding;

(4) 3j+2 and 3j+3 two field pictures carry out lossless coding between space-time many predictive frame with reference to 3j+1 two field picture; Comprise following sub-step:

(4.1) at reference frame f _3j+1middle search current sub-block f _c,i, c=3j+2 or 3j+3; I=1,2 ..., the best matching blocks of SumI, obtains motion vector (m ₀, n ₀);

(4.2) according to above-mentioned motion vector (m ₀, n ₀), to current sub-block f _c,iin each pixel f _c,i(m, n) utilizes three fundamental forecasting devices to predict respectively, obtains the pixel value predicted d ∈ [1,2,3];

(4.3) current sub-block f is calculated _c,iin the accumulation predicated error absolute value SAD of all pixels ^d:

${\mathrm{SAD}}^d=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}|{\stackrel{\&OverBar;}{x}}_k^d(m,n)-x_k(m,n)|,d\&Element;\left[\mathrm{1,2,3}\right]$

Wherein, x _k(m, n)=f _c,i(m, n)

(4.4) select the fallout predictor making accumulation predicated error absolute value minimum as the fixing fallout predictor of this sub-block:

$\mathrm{mode}=\arg \underset{d=\mathrm{1,2,3}}{\min }\left\{{\mathrm{SAD}}^d\right\}$

(4.5) utilize fixing fallout predictor that step (4.4) is tried to achieve to current sub-block f _c,iin each pixel f _c,i(m, n) predicts, then adopts based on contextual Golomb entropy code predicated error, thus obtains compressed bit stream.

Further, three fundamental forecasting implement bodies in described step (4.2) are:

(4.2.1) fallout predictor 1: the Pixel Information of reference frame do not considered by this fallout predictor, only utilizes the neighbor in present frame to predict:

${\stackrel{\&OverBar;}{x}}_k^1=a_k+b_k-c_k$ Other

Wherein T ₁∈ [10,20] and T ₂∈ [10,20] is respectively threshold value;

(4.2.2) fallout predictor 2: the pixel in reference frame used by this fallout predictor, utilizes motion vector (m ₀, n ₀) neighborhood pixels that obtains predicts this pixel, finally linear weighted function correction again, that is:

$x_k^1=x_{k-1}+b_k-b_{k-1}$

$x_k^2=x_{k-1}+a_k-a_{k-1}$

$x_k^3=a_k+b_k-c_{k-1}$

By the mean value of above-mentioned three predicted values as predicting the outcome:

(4.2.3) fallout predictor 3: in adjacent two two field pictures of this fallout predictor hypothesis, closely, so utilize the predicated error of respective pixel in reference frame to predict current pixel, concrete form is the predicated error of respective pixel:

$\stackrel{\&OverBar;}{x}=a_{k-1}+b_{k-1}-c_{k-1}$

$\stackrel{\&OverBar;}{x}k=a_k+b_k-c_k$

Wherein pixel each in above-mentioned prediction module is defined as:

a _k-1＝f _3j+1(m-m ₀,n-n ₀-1) a _k＝f _c,i(m,n-1)

b _k-1＝f _3j+1(m-m ₀-1,n-n ₀) b _k＝f _c,i(m-1,n)

c _k-1＝f _3j+1(m-m ₀-1,n-n ₀-1) c _k＝f _c,i(m-1,n-1)

d _k-1＝f _3j+1(m-m ₀-1,n-n ₀+1) d _k＝f _c,i(m-1,n+1)

x _k-1＝f _3j+1(m-m ₀,n-n ₀) x _k＝f _c,i(m,n)。

According to another aspect of the present invention, additionally provide a kind of Lossless Compression system based on the many predictive modes of space-time, comprise as lower module:

Sequence image acquisition module, obtains sequence image f for utilizing spaceborne imaging system;

Partition module, for each two field picture f that sequence image acquisition module is obtained _k, k=1,2, be divided into non-overlapping copies and size is the sub-block f of M × N _k,i, i=1,2 ..., SumI, SumI are sub-block sum, and M, N are preset value;

JPEG-LS coding module, for the 3j+1 by partition CMOS macro cell, j=0,1 ... two field picture is reference frame image, carries out piecemeal JPEG-LS coding;

The interframe lossless coding of many predictive modes, for carrying out lossless coding between space-time many predictive frame to 3j+2 and 3j+3 two field picture with reference to 3j+1 two field picture; Comprise following submodule:

Motion vector obtains submodule, at reference frame f _3j+1middle search current sub-block f _c,i, c=3j+2 or 3j+3; I=1,2 ..., the best matching blocks of SumI, obtains motion vector (m ₀, n ₀);

Fundamental forecasting device predictor module, for current sub-block f _c,iin each pixel f _c,i(m, n) utilizes three fundamental forecasting devices to predict respectively;

Error Absolute Value calculating sub module, for calculating current sub-block f _c,iin the accumulation predicated error absolute value SAD of all pixels ^d:

${\mathrm{SAD}}^d=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}|{\stackrel{\&OverBar;}{x}}_k^d(m,n)-x_k(m,n)|,d\&Element;\left[\mathrm{1,2,3}\right]$

Fallout predictor chooser module, for the fallout predictor selecting the to make accumulation predicated error absolute value minimum fixing fallout predictor as this sub-block:

$\mathrm{mode}=\arg \underset{d=\mathrm{1,2,3}}{\min }\left\{{\mathrm{SAD}}^d\right\}$

Encoding submodule, for utilizing the fixing fallout predictor of fallout predictor chooser model choice to current sub-block f _c,iin each pixel f _c,i(m, n) predicts, then adopts based on contextual Golomb entropy code predicated error, thus obtains compressed bit stream.

Further, described fundamental forecasting device predictor module specifically comprises:

Fallout predictor 1: the Pixel Information of reference frame do not considered by this fallout predictor, only utilizes the neighbor in present frame to predict:

${\stackrel{\&OverBar;}{x}}_k^1=a_k+b_k-c_k$ Other

Wherein T ₁∈ [10,20] and T ₂∈ [10,20] is respectively threshold value;

Fallout predictor 2: the pixel in reference frame used by this fallout predictor, utilizes motion vector (m ₀, n ₀) neighborhood pixels that obtains predicts this pixel, finally linear weighted function correction again, that is:

$x_k^1=x_{k-1}+b_k-b_{k-1}$

$x_k^2=x_{k-1}+a_k-a_{k-1}$

$x_k^3=a_k+b_k-c_{k-1}$

By the mean value of above-mentioned three predicted values as predicting the outcome:

Fallout predictor 3: in adjacent two two field pictures of this fallout predictor hypothesis, closely, so utilize the predicated error of respective pixel in reference frame to predict current pixel, concrete form is the predicated error of respective pixel:

$\stackrel{\&OverBar;}{x}=a_{k-1}+b_{k-1}-c_{k-1}$

$\stackrel{\&OverBar;}{x}k=a_k+b_k-c_k$

Wherein, pixel each in above-mentioned prediction module is defined as:

a _k-1＝f _3j+1(m-m ₀,n-n ₀-1) a _k＝f _c,i(m,n-1)

b _k-1＝f _3j+1(m-m ₀-1,n-n ₀) b _k＝f _c,i(m-1,n)

c _k-1＝f _3j+1(m-m ₀-1,n-n ₀-1) c _k＝f _c,i(m-1,n-1)

d _k-1＝f _3j+1(m-m ₀-1,n-n ₀+1) d _k＝f _c,i(m-1,n+1)

x _k-1＝f _3j+1(m-m ₀,n-n ₀) x _k＝f _c,i(m,n)。

Conventional compression method JPEG-LS just utilizes image space correlation, and inter-frame difference and the interframe prediction encoding method with motion compensation just utilize sequence image correlation in time, and therefore the compression effectiveness of these compression methods to sequence image is bad.Lossless compression method based on the many predictive modes of space-time proposed by the invention combines the correlation of image on room and time and improves prediction mode, image is divided into some sub-blocks, the adaptively selected optimum prediction mode of different sub-block is predicted, thus making the region of fallout predictor to image different characteristic have adaptivity, the compression effectiveness of lossless coding method to sequence image therefore based on the many predictive modes of space-time is better.

Accompanying drawing explanation

Fig. 1 is the exemplary plot that inter-frame difference prediction was lost efficacy:

Fig. 2 is the flow chart of the interframe decoding method with motion compensation;

Fig. 3 is estimation basic principle schematic;

Fig. 4 is the prediction module of the lossless compression method that the present invention is based on the many predictive modes of space-time;

Fig. 5 is cycle tests image;

Fig. 5 (a)-(c) is the satellite sequence image first frame example of 10bits;

Fig. 5 (d)-(f) is the satellite sequence image first frame example of 12bits.

Fig. 5 (g) is the image first frame example of taking a flight test of 8bits;

Fig. 5 (h) is the parking lot image first frame example of 8bits;

Fig. 6 is the satellite sequence image that Fig. 5 (a) is complete;

Fig. 7 (a) is JPEG-LS method, have the inter-frame encoding methods of motion compensation, LOCO-3D and based on the lossless compression method of space-time many predictive modes, sequential test image graph 5 is divided into the average compression block diagram of 16 × 16 sub-blocks;

Fig. 7 (b) is JPEG-LS method, have the inter-frame encoding methods of motion compensation, LOCO-3D and based on the lossless compression method of space-time many predictive modes, sequential test image graph 5 is divided into the average compression block diagram of 16 × 32 sub-blocks;

Proportion map when Fig. 8 carries out space-time many predictions Lossless Compression the sub-block that image is divided into 16 × 16 shared by three fundamental forecasting devices;

Fig. 9 is the lossless compression method schematic diagram that the present invention is based on the many predictive modes of space-time.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.In addition, if below in described each execution mode of the present invention involved technical characteristic do not form conflict each other and just can mutually combine.

As shown in Figure 9, the invention provides a kind of lossless compression method based on the many predictive modes of space-time, its step comprises:

(1) spaceborne imaging system is utilized to obtain sequence image f;

(2) each two field picture f that step (1) obtains _k, k=1,2 ... be divided into non-overlapping copies and size is the sub-block f of M × N _k,i, i=1,2 ..., SumI, SumI are sub-block sum, and M, N are preset value;

(3) 3j+1, j=0 in step (2) is made, 1 ... two field picture is reference frame image, carries out piecemeal JPEG-LS coding;

(4) 3j+2 and 3j+3 two field pictures carry out lossless coding between space-time many predictive frame with reference to 3j+1 two field picture:

(4.1) at reference frame f _3j+1middle search current sub-block f _c,i, c=3j+2 or 3j+3i; The best matching blocks of=1,2..., SumI, obtains motion vector (m ₀, n ₀);

(4.2) to current sub-block f _c,iin each pixel f _c,i(m, n) utilizes three fundamental forecasting devices to predict respectively, first introduces some basic definitions below, is illustrated in figure 4 the definition of each pixel in the prediction module of the lossless compression method that the present invention is based on the many predictive modes of space-time:

a _k-1＝f _3j+1(m-m ₀,n-n ₀-1) a _k＝f _c,i(m,n-1)

b _k-1＝f _3j+1(m-m ₀-1,n-n ₀) b _k＝f _c,i(m-1,n)

c _k-1＝f _3j+1(m-m ₀-1,n-n ₀-1) c _k＝f _c,i(m-1,n-1)

d _k-1＝f _3j+1(m-m ₀-1,n-n ₀+1) d _k＝f _c,i(m-1,n+1)

x _k-1＝f _3j+1(m-m ₀,n-n ₀) x _k＝f _c,i(m,n)

(4.2.1) fallout predictor 1: the Pixel Information of reference frame do not considered by this fallout predictor, only utilizes the neighbor in present frame to predict:

${\stackrel{\&OverBar;}{x}}_k^1=a_k+b_k-c_k$ Other

Wherein T ₁∈ [10,20] and T ₂∈ [10,20] is respectively threshold value.Fallout predictor 1 changes greatly mainly for image scene or target moves the situation causing frame-to-frame correlation less.

(4.2.2) fallout predictor 2: the pixel in reference frame used by this fallout predictor, utilizes motion vector (m ₀, n ₀) neighborhood pixels that obtains predicts this pixel, finally linear weighted function correction again, that is:

$x_k^1=x_{k-1}+b_k-b_{k-1}$

$x_k^2=x_{k-1}+a_k-a_{k-1}$ $x_k^3=a_k+b_k-c_{k-1}$

By the mean value of above-mentioned three predicted values as predicting the outcome:

(4.2.3) fallout predictor 3: in adjacent two two field pictures of this fallout predictor hypothesis, closely, so utilize the predicated error of respective pixel in reference frame to predict current pixel, concrete form is the predicated error of respective pixel:

$\stackrel{\&OverBar;}{x}=a_{k-1}+b_{k-1}-c_{k-1}$

$\stackrel{\&OverBar;}{x}k=a_k+b_k-c_k$

(4.3) current sub-block f is calculated _c,iin the accumulation predicated error absolute value SAD of all pixels ^d:

${\mathrm{SAD}}^d=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}|{\stackrel{\&OverBar;}{x}}_k^d(m,n)-x_k(m,n)|,d\&Element;\left[\mathrm{1,2,3}\right]$

(4.4) select the fallout predictor making accumulation predicated error absolute value minimum as the fixing fallout predictor of this sub-block:

$\mathrm{mode}=\arg \underset{d=\mathrm{1,2,3}}{\min }\left\{{\mathrm{SAD}}^d\right\}$

(4.5) utilize fixing fallout predictor that step (4.4) is tried to achieve to current sub-block f _c,iin each pixel f _c,i(m, n) predicts, then adopts based on contextual Golomb entropy code predicated error, thus obtains compressed bit stream.

Carry out Lossless Compression contrast experiment to the cycle tests image in Fig. 5, its result is as shown in Fig. 7, table 1 and table 2, and the method that experimental result reflects the present invention's proposition utilizes the correlation of sequence satellite image on room and time, so obtain higher compression ratio; And along with the increase of coding piecemeal, fewer owing to preserving the pixel that sub-block selects to fill when the information of fundamental forecasting device and coding, therefore compression ratio is higher.LOCO-3D is (see document: D.Brunello, G.Calvagno, G.Mian and R.Rinaldo.Lossless Compression of Video Using Temporal Information.IEEETransactions on Image Processing, Vol.12, No.2, pp.132 ~ 139,2003.) although the method that proposes of compression effectiveness and the present invention close, the prediction complexity of the method that its prediction complexity proposes than the present invention is high a lot.

Table 1 is JPEG-LS method, have the inter-frame encoding methods of motion compensation, LOCO-3D and based on the lossless compression method of space-time many predictive modes, sequential test image graph 5 is divided into 16 × 16 compression ratio.

Table 1

Table 2 is JPEG-LS methods, have the inter-frame encoding methods of motion compensation, LOCO-3D and based on the lossless compression method of space-time many predictive modes, sequential test image graph 5 is divided into 16 × 32 compression ratio.

Table 2

The method that experimental result shown in accompanying drawing 8 and table 3 reflects the present invention's proposition can select optimum predictive mode adaptively according to the actual conditions of current sub-block in three fundamental forecasting devices, thus reaches best encoding efficiency.

Table 3 is sub-block that image is divided into 16 × 16 numbers when carrying out space-time many predictions Lossless Compression shared by three fundamental forecasting devices.

Table 3

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (4)

1. based on a lossless compression method for the many predictive modes of space-time, it is characterized in that, described method comprises the steps:

(1) spaceborne imaging system is utilized to obtain sequence image f;

(2) each two field picture f that step (1) obtains _k, k=1,2 ... be divided into non-overlapping copies and size is the sub-block f of M × N _k,i, i=1,2 ..., SumI, SumI are sub-block sum, and M, N are preset value;

(3) 3j+1, j=0 in step (2) is made, 1 ... two field picture is reference frame image, carries out piecemeal JPEG-LS coding;

(4) 3j+2 and 3j+3 two field pictures carry out lossless coding between space-time many predictive frame with reference to 3j+1 two field picture; Comprise following sub-step:

(4.1) at reference frame f _3j+1middle search current sub-block f _c,i, c=3j+2 or 3j+3; I=1,2 ..., the best matching blocks of SumI, obtains motion vector (m ₀, n ₀);

(4.2) according to above-mentioned motion vector (m ₀, n ₀), to current sub-block f _c,iin each pixel f _c,i(m, n) utilizes three fundamental forecasting devices to predict respectively, obtains the pixel value predicted

(4.3) current sub-block f is calculated _c,iin the accumulation predicated error absolute value SAD of all pixels ^d:

${\mathrm{SAD}}^d=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}|{\stackrel{\&OverBar;}{x}}_k^d(m,n)-x_k(m,n)|,d\&Element;\left[\mathrm{1,2,3}\right]$

Wherein, x _k(m, n)=f _c,i(m, n)

(4.4) select the fallout predictor making accumulation predicated error absolute value minimum as the fixing fallout predictor of this sub-block:

$\mathrm{mode}=\arg \underset{d=\mathrm{1,2,3}}{\min }\left\{{\mathrm{SAD}}^d\right\}$

(4.5) utilize fixing fallout predictor that step (4.4) is tried to achieve to current sub-block f _c,iin each pixel f _c,i(m, n) predicts, then adopts based on contextual Golomb entropy code predicated error, thus obtains compressed bit stream.

2. the method for claim 1, is characterized in that, three fundamental forecasting implement bodies in described step (4.2) are:

(4.2.1) fallout predictor 1: the Pixel Information of reference frame do not considered by this fallout predictor, only utilizes the neighbor in present frame to predict:

other

Wherein T ₁∈ [10,20] and T ₂∈ [10,20] is respectively threshold value;

(4.2.2) fallout predictor 2: the pixel in reference frame used by this fallout predictor, utilizes motion vector (m ₀, n ₀) neighborhood pixels that obtains predicts this pixel, finally linear weighted function correction again, that is:

$x_k^1=x_{k-1}+b_k-b_{k-1}$

$x_k^2=x_{k-1}+a_k-a_{k-1}$

$x_k^3=a_k+b_k-c_{k-1}$

By the mean value of above-mentioned three predicted values as predicting the outcome:

(4.2.3) fallout predictor 3: in adjacent two two field pictures of this fallout predictor hypothesis, closely, so utilize the predicated error of respective pixel in reference frame to predict current pixel, concrete form is the predicated error of respective pixel:

$\stackrel{\&OverBar;}{x}=a_{k-1}+b_{k-1}-c_{k-1}$

${\stackrel{\&OverBar;}{x}}_k=a_k+b_k-c_k$

Wherein pixel each in above-mentioned prediction module is defined as:

a _k-1＝f _3j+1(m-m ₀,n-n ₀-1) a _k＝f _c,i(m,n-1)

b _k-1＝f _3j+1(m-m ₀-1,n-n ₀) b _k＝f _c,i(m-1,n)

c _k-1＝f _3j+1(m-m ₀-1,n-n ₀-1) c _k＝f _c,i(m-1,n-1)

d _k-1＝f _3j+1(m-m ₀-1,n-n ₀+1) d _k＝f _c,i(m-1,n+1)

x _k-1＝f _3j+1(m-m ₀,n-n ₀) x _k＝f _c,i(m,n)。

3. based on a Lossless Compression system for the many predictive modes of space-time, it is characterized in that, described system comprises as lower module:

Sequence image acquisition module, obtains sequence image f for utilizing spaceborne imaging system;

Partition module, for each two field picture f that sequence image acquisition module is obtained _k, k=1,2, be divided into non-overlapping copies and size is the sub-block f of M × N _k,i, i=1,2 ..., SumI, SumI are sub-block sum, and M, N are preset value;

JPEG-LS coding module, for the 3j+1 by partition CMOS macro cell, j=0,1 ... two field picture is reference frame image, carries out piecemeal JPEG-LS coding;

The interframe lossless coding of many predictive modes, for carrying out lossless coding between space-time many predictive frame to 3j+2 and 3j+3 two field picture with reference to 3j+1 two field picture, comprises following submodule:

Motion vector obtains submodule, at reference frame f _3j+1middle search current sub-block f _c,i, c=3j+2 or 3j+3; I=1,2 ..., the best matching blocks of SumI, obtains motion vector (m ₀, n ₀);

Fundamental forecasting device predictor module, for current sub-block f _c,iin each pixel f _c,i(m, n) utilizes three fundamental forecasting devices to predict respectively;

Error Absolute Value calculating sub module, for calculating current sub-block f _c,iin the accumulation predicated error absolute value SAD of all pixels ^d:

${\mathrm{SAD}}^d=\underset{m=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}|{\stackrel{\&OverBar;}{x}}_k^d(m,n)-x_k(m,n)|,d\&Element;\left[\mathrm{1,2,3}\right]$

Fallout predictor chooser module, for the fallout predictor selecting the to make accumulation predicated error absolute value minimum fixing fallout predictor as this sub-block:

$\mathrm{mode}=\arg \underset{d=\mathrm{1,2,3}}{\min }\left\{{\mathrm{SAD}}^d\right\}$

Encoding submodule, for utilizing the fixing fallout predictor of fallout predictor chooser model choice to current sub-block f _c,iin each pixel f _c,i(m, n) predicts, then adopts based on contextual Golomb entropy code predicated error, thus obtains compressed bit stream.

4. method as claimed in claim 3, it is characterized in that, described fundamental forecasting device predictor module specifically comprises:

Fallout predictor 1: the Pixel Information of reference frame do not considered by this fallout predictor, only utilizes the neighbor in present frame to predict:

other

Wherein T ₁∈ [10,20] and T ₂∈ [10,20] is respectively threshold value;

Fallout predictor 2: the pixel in reference frame used by this fallout predictor, utilizes motion vector (m ₀, n ₀) neighborhood pixels that obtains predicts this pixel, finally linear weighted function correction again, that is:

$x_k^1=x_{k-1}+b_k-b_{k-1}$

$x_k^2=x_{k-1}+a_k-a_{k-1}$

$x_k^3=a_k+b_k-c_{k-1}$

By the mean value of above-mentioned three predicted values as predicting the outcome:

Fallout predictor 3: in adjacent two two field pictures of this fallout predictor hypothesis, closely, so utilize the predicated error of respective pixel in reference frame to predict current pixel, concrete form is the predicated error of respective pixel:

$\stackrel{\&OverBar;}{x}=a_{k-1}+b_{k-1}-c_{k-1}$

${\stackrel{\&OverBar;}{x}}_k=a_k+b_k-c_k$

Wherein, pixel each in above-mentioned prediction module is defined as:

a _k-1＝f _3j+1(m-m ₀,n-n ₀-1) a _k＝f _c,i(m,n-1)

b _k-1＝f _3j+1(m-m ₀-1,n-n ₀) b _k＝f _c,i(m-1,n)

c _k-1＝f _3j+1(m-m ₀-1,n-n ₀-1) c _k＝f _c,i(m-1,n-1)

d _k-1＝f _3j+1(m-m ₀-1,n-n ₀+1) d _k＝f _c,i(m-1,n+1)

x _k-1＝f _3j+1(m-m ₀,n-n ₀) x _k＝f _c,i(m,n)。