CN102088602A - Code rate control method for JPEG-LS (joint photographic experts group-lossless standard) image compression - Google Patents

Abstract

The invention provides a code rate control method for JPEG-LS (joint photographic experts group-lossless standard) image compression, which is used to dynamically adjust a parameter NEAR value based on the accumulated deviation value of an actual code rate and a target code rate, thus ensuring that different image contents can be output at a code rate approximate to a required code rate after being subjected to JPEG-LS compression. The method is fast in convergence rate and is easy to implement, can accurately control the JPEG-LS output code rate, and can ensure preferable compression performance. According to the code rate control method, a corresponding hardware implementation mode is designed. A compression encoding module and a code rate control module form a feedback loop; the feedback loop adaptively adjusts the parameter NEAR value for segmented sub-images; and the output compressed code stream is cached through a cache control module at first, and is then output at a constant code rate. Tests prove that the hardware is simple and effective to implement, is stable in control, and can ensure that the recovered images have good quality.

Description

A kind of bit rate control method of JPEG-LS image compression

Technical field

The present invention relates to a kind of bit rate control method of JPEG-LS image compression, belong to the digital image compression field.

Background technology

The lossless data compression standard of generally acknowledging function admirable at present in the world is the JPEG-LS algorithm, but the output code flow of this algorithm has bigger fluctuation, and code check is uncontrollable, and this has just limited its application in practice.Especially in remote sensing satellite was used, channel width was subjected to strict restriction, required the bit rate output of encoder to be complementary with channel code rate, otherwise bit rate output is too high, can cause data to overflow, and causes information dropout, bit rate output is low excessively, can cause the waste of channel resource again.Therefore the rate control algorithm of studying the JPEG-LS image compression has very important realistic meaning.

The Rate Control problem is not strange in the image compression field, in field of video compression, rate control algorithm has obtained using widely, the most typical rate control algorithm that just is based on the rate distortion analysis, it not only can well regulate the bit rate output after the compression, and the quality of assurance video image that can be high as far as possible, now be applied to H.264, in the video image compression standards such as MPEG4.In Joint Photographic Experts Group, difference according to Control Parameter, it can be divided into two big classes, the first kind is exactly the compression standard of control code rate, such as JPEG2000, and CCSDS, compression algorithms such as SPITE, this class algorithm all is based on the compression algorithm of small echo, have the function that self adaptation is adjusted bit rate output, but reconstructed image quality is uncontrollable; Second class is exactly the compression standard of control quality, and as the JPEG-LS compression standard, it can directly control the recovery quality of reconstructed image, but because this algorithm adopts based on contextual predictive coding mode, its bit rate output is uncontrollable.According to different compression standards, usually need to adopt different Rate Control strategies, at present, rate control algorithm research for the JPEG-LS compression is also few, the single order bit rate control method of mentioning in " jpeg-ls image compression dynamic code rate control strategy " that existing Xu Yan insults adopts part to regulate array value and fixed gate limit value Th to NEAR, therefore it is relatively poor to control effect, and the second order bit rate control method calculation of complex of mentioning, be unfavorable for the hardware realization, " Open-loop rate controlfor JPEC-LS near lossless image compression " every row that external J.Jiang and M.Reddy proposes all will carry out the adjustment of a NEAR value, and adjustment process is comparatively complicated, though the control precision height, bigger loss the compression performance of image.

Summary of the invention

Technology of the present invention is dealt with problems and is: overcome the deficiencies in the prior art, at the uncontrollable defective of JPEG-LS Standard of image compression bit rate output, the present invention proposes a kind of JPEG-LS image compression code check control method, not only can realize the control of target bit rate more accurately, fast convergence rate, be easy to realize, and can also guarantee compression performance preferably.

Technical scheme of the present invention is as follows:

A kind of bit rate control method of JPEG-LS image compression, step is as follows:

(1) the JPEG-LS image with input carries out the subgraph division according to same size, and the subgraph size is r*c, and the pixel that r is capable and c is listed as is promptly arranged, and the subgraph size satisfies

Corresponding one of each subgraph is dynamically adjusted factor NEAR, and each subgraph all is independently processing units; Described dynamic adjustment factor NEAR is the worst error that the picture quality recovery is allowed;

(2) determine dynamically initial value initial_NEAR, minimum M in_NEAR and the maximum Max_NEAR of adjustment factor NEAR according to the JPEG-LS image of target compression ratio and input; After dynamically the minimum M in_NEAR of adjustment factor NEAR and maximum Max_NEAR decide, in the process that the JPEG-LS image of importing compresses, be fixed constant;

(3) according to the dynamic adjustment factor NEAR that obtains, current subgraph is carried out the JPEG-LS compressed encoding, calculate the cumulative departure amount E (i) of target compression ratio and current compression ratio;

(4) determine dynamically to adjust compare threshold t1 and t2 by following formula:

$\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="HSA00000405497200011.png,HSA00000405497200031.png"\; state="\{\{state\}\}">t</figure-callout>}1=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{CCR}(i-t)/\mathrm{CR}*\mathrm{\&alpha;},$ $\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="HSA00000405497200011.png,HSA00000405497200031.png"\; state="\{\{state\}\}">t</figure-callout>}2=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{CCR}(i-t)/\mathrm{CR}*\mathrm{\&beta;},$

Wherein, CR is a target compression ratio, and α, β are thresholding coefficient and α=1/8, β=1/32, and CCR (i-t) is the actual compression code flow of i-t subgraph, and when t=0, CCR (i) is the actual compression code flow of i subgraph, and i is a nonnegative integer; N dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, and n is nonnegative integer, and initial value is 0;

(5) according to the t1, the t2 that obtain in cumulative departure amount E (i) that obtains in the step (3) and the step (4), pass through formula

Calculate the updating value of dynamically adjusting factor NEAR;

Wherein, Δ Q is the adjustment step-length, and

NEAR (i) is the value of the dynamic adjustment factor NEAR that gets of i subgraph, NEAR (i-1) is a last subgraph, the i.e. value of the dynamic adjustment factor NEAR that gets of i-1 subgraph, initial_NEAR is an initial value of dynamically adjusting factor NEAR, n dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, N is the accumulative total frequency

And N 〉=16, wherein, R*C represent the size of continuously adjustable input picture be R capable * C row, R=4096, the fabric width of C presentation video are the C row, and be consistent with the fabric width of the JPEG-LS image of input described in the step (1);

(6) the value NEAR (i) of the dynamic adjustment factor NEAR that gets according to i the subgraph that obtains in the step (5) carries out clamper by following formula to NEAR (i), enters step (7) afterwards;

$\mathrm{NEAR}\left(i\right)=\left\{\begin{array}{c}\mathrm{Max}\_\mathrm{NEAR},\mathrm{NEAR}\left(i\right)>\mathrm{Max}\_\mathrm{NEAR}\\ \mathrm{Min}\_\mathrm{NEAR},\mathrm{NEAR}\left(i\right)<\mathrm{Min}\_\mathrm{NEAR}\end{array}\right.,$

Wherein, Min_NEAR is a minimum value of dynamically adjusting factor NEAR, and Max_NEAR is a maximum;

(7) carry out following operation to adjusting frequency n: give n with the n+1 assignment, promptly n=n+1 enters step (8) afterwards;

(8) judge whether n equals N,, forward step (2) to, if n ≠ N then forwards step (3) to if n=N then makes n=0.

The cumulative departure amount E (i) that calculates target compression ratio and current compression ratio described in the step (3) is undertaken by following formula:

$E\left(i\right)=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}|\mathrm{OCR}-\mathrm{CCR}(i-t)|,n=\{\mathrm{0,1},...,N-1\};$

Wherein, OCR represents the targeted compression code flow, and N is the accumulative total frequency.

The present invention's advantage compared with prior art is:

(1) used Rate Control to adjust strategy continuously among the present invention, this is mainly reflected in two aspects: the one, for the calculating of the departure E (i) of target compression ratio and current compression ratio, this value is an accumulated value in the continuous adjustment cycle N (accumulative total frequency), is different from general deviate and calculates; The 2nd, updating value NEAR follows the example of, and this value is non return to zero in continuous adjustment cycle N (accumulative total frequency), promptly is a continuously adjustable process.This continuously adjustable strategy has not only solved when the texture sudden change takes place in the image, adjust that the factor can not be followed the tracks of the epigraph content change immediately and the Real Time Compression that causes than the situation of discontented foot-eye compression ratio, and remedied and taken into account less stress or the toning problem that compression performance brings in the rate control algorithm, therefore, this adjustment strategy makes that not only the present invention has wide range of applications, and control performance is stable.

(2) mentioned the problem of following the example of of dynamic adjustment factor initial value among the present invention, the initial value that should dynamically adjust the factor is according to the target compression ratio of the characteristics (as precision) of image to be adjusted and requirement and different, be different from general the following the example of of a fixed value of getting, so just can accelerate the convergence rate of rate control algorithm so that adjusting the initial value of the factor drops near the target value of image in most cases; After adjusting factor adaptive simultaneously and adjusting a period of time, exchange integral divisor initialize again, with the solution inaccurate problem of control that correlation is brought after descending between the subgraph far away of being separated by.

(3) mentioned the problem of the adjustment factor after upgrading being carried out clamper among the present invention, if the scope that the NEAR value is adjusted does not add restriction, can make that the fluctuation range of search NEAR value is excessive, the speed of tracking image information is slow excessively, thus can't response diagram as the truth of information change.When doing the dynamic adjustment of NEAR value, maximum to NEAR limits, make the NEAR value of complicated image obtain less than normal, simultaneously the minimum value of NEAR is carried out clamper, make the NEAR value of simple image obtain bigger than normal, and the NEAR value is big more, and the degree that influence recovers picture quality is serious more, and the clamper operation of therefore adjusting the factor has also guaranteed to recover preferably picture quality to a certain extent.

(4) mentioned dynamic adjustment compare threshold t1 among the present invention, the problem of following the example of of t2, this compare threshold is not only relevant with the compressed capability of conventional images, and it is relevant with target compression ratio, also dynamic change in the adjustment process of whole Rate Control, this threshold value follow the example of and fixedly following the example of of existing threshold value compared, have stronger adaptive adjustment capability.

(5) mentioned the problem of input picture being carried out the subgraph division among the present invention, divide by input picture being carried out subgraph, make the identical NEAR value of the interior employing of same subgraph compress, the compression performance that as far as possible keeps JPEG-LS, different subgraphs then adopts the NEAR value after the renewal to compress, to reach the purpose of the output code flow of controlling compressed bit stream.The size of subgraph r*c satisfies

The compression performance of JPEG-LS will be kept greatly, because the compression performance of JPEG-LS is more responsive to fabric width c.

(6) mentioned following the example of of adjustment step delta Q among the present invention, variable quantity for Δ Q only defines three kinds of situations: 0,1 and 2, this following the example of not only makes hardware realize simply also meeting the universal law of image change, also guaranteed certain recovery picture quality simultaneously.Otherwise if the variable quantity scope of Δ Q is excessive, branch is too much, not only hardware is realized complicated, and it is big to cause real-time output code flow to fluctuate owing to the amplitude of adjusting the step-length adjustment is excessive, and the NEAR value difference is different greatly between the adjacent subgraph of recovery image, recovers the picture quality reduction thereby cause.

Description of drawings

Fig. 1 is the compression performance curves of remote sensing 10 bit image under different N EAR value;

Fig. 2 uses a condensing encoder of the present invention to realize block diagram;

Fig. 3 is the compression performance curves of remote sensing 8 bit image under different N EAR value;

Fig. 4 is that hardware of the present invention is realized port definition;

Fig. 5 is the Rate Control result of the present invention under remote sensing images 4 multiplication of voltages contract;

Fig. 6 is a flow chart of the present invention.

Embodiment

The specific implementation of the rate control algorithm of JPEG-LS compression is described below in conjunction with accompanying drawing and specific implementation example:

Fig. 1 is the more representational 10 bit remote sensing images of 9 amplitude ratios, and after the value in NEAR value traversal [0,9] scope, the compression performance curve that obtains, abscissa are the NEAR value, and ordinate is the compression ratio that obtains, and the image size is 2048*2048.Symbol among the figure " alashan ", " baoding ", " ice ", " india ", " ocean ", " pinshan ", " qinhuangdao ", " shijiazhuang ", " simao " only are label, do not have physical meaning.Wherein, " india " corresponding curve 1, " ocean " corresponding curve 2, " ice " corresponding curve 3, " pinshan " corresponding curve 4, " alashan " corresponding curve 5, " baoding " corresponding curve 6, " qinhuangdao " corresponding curve 7, " simao " corresponding curve 8, " shijiazhuang " corresponding curve 9.

According to this width of cloth curve chart, we can obtain some compression rules of such image, and when rate control algorithm of the present invention was applied to the remote sensing images field, the value of some constant served as to obtain according to summing up with this curve chart.As being that 2 multiplication of voltages are when contracting when target compression ratio, the straight line (shown in the dotted line among the figure " 2 multiplication of voltages contract ") of we the first CompressRatio=2 that draws in the drawings, the abscissa value that this dotted line is corresponding with every intersections of complex curve is the contract value of corresponding parameters NEAR of 2 multiplication of voltages, value as the corresponding abscissa NEAR of the intersection point of curve shijiazhuang (curve 9) is 1, the intersection point of curve india (curve 1) and 2 multiplication of voltage diminishing line correspondences is on the reverse extending line of curve, show when NEAR gets minimum value 0, the compression ratio of india is greater than 2 times, the value of the corresponding abscissa NEAR of the intersection point of curve qinhuangdao (curve 7) is between 0～1, relatively approach 1, or the like, value according to the corresponding NEAR of all curves distributes, as can be seen, when 2 multiplication of voltages contract, 9 width of cloth remote sensing images among Fig. 1 are got NEAR just can reach or exceed 2 times compression ratio at 1 o'clock, therefore getting initial_NEAR=1, to carry out the code check adjustment that 2 multiplication of voltages contract be only, it is best that the minimum value of NEAR is taken as Min_NEAR=0, because in JPEG-LS compressed encoding standard, regulation NEAR is a nonnegative integer, and as can be seen from Figure 1, there are a few width of cloth images when NEAR=0, just can reach the compression ratio that 2 multiplication of voltages contract, it is best that the maximum of NEAR is taken as Max_NEAR=2, although it is 1 just passable that the NEAR value maximum that 9 width of cloth remote sensing images, 2 multiplication of voltages of enumerating in Fig. 1 contract is got, the span of NEAR itself is just little but 2 multiplication of voltages contract, therefore the span that can suitably relax some NEAR makes the accommodation of algorithm wider.As target compression ratio is 4 multiplication of voltages when contracting, parameter initial_NEAR, the obtaining value method of Min_NEAR and Max_NEAR and above be similar.In summary, it is exactly characteristic according to input picture, as the pixel precision, spatial resolution etc., select such input picture and have the different images content (as land, the harbour, city etc.) a few width of cloth representative image, carry out the JPEG-LS compression of NEAR value traversal earlier, the curve relation figure of the NEAR value of drawing and compression ratio CompressRatio, again according to target compression ratio, find out the NEAR value of different representative image correspondence under this compression ratio, according to these NEAR values, getting initial_NEAR is the median of concentrated NEAR value scope of value, like this in follow-up adjustment process, the NEAR value will increase or reduce all is the fastest, minimum value and maximum in Min_NEAR and the corresponding value of Max_NEAR difference, certainly, Min_NEAR and Max_NEAR also can adjust a bit on this basis a little, to adapt to different concrete conditions.

Fig. 2 is the more representational 8 bit remote sensing images of 12 amplitude ratios, and after the value in NEAR value traversal [0,11] scope, the compression performance curve that obtains, abscissa are the NEAR value, and ordinate is the compression ratio that obtains, and the image size is 1024*1024.Similar with Fig. 1, the purpose of making this width of cloth statistical chart also is to be used for instructing the value of such image at the corresponding constant of rate control algorithm.

Fig. 3 is a realization block diagram of using a JPEG-LS condensing encoder of the present invention, below in conjunction with this block diagram, carries out 4 multiplication of voltages with remote sensing 10 bit image and is condensed to example, introduces the specific implementation of rate control algorithm in detail.

One, image buffer storage module

This module mainly is to finish original input picture is carried out image division according to the requirement of hardware realization condition and rate control algorithm, the input mode of general remote sensing image data is exactly along with pushing away line by line of CCD camera sweeps into the row input of pixel line by line, except the fabric width of image is determined by the parameter of CCD camera, input picture does not have limitation in height, in order to carry out Rate Control, and make things convenient for the parallel processing of hardware and improve the antijamming capability of encoding, must carry out subgraph to the remote sensing input picture divides, it is fixed that the size of subgraph will be come according to the size of input picture, if input picture is less, then in order to guarantee to adjust also corresponding get less of number of times subgraph, otherwise if input picture is bigger, subgraph can correspondingly be obtained big to keep the compression performance of JPEG-LS preferably.After subgraph was divided, each subgraph adopted unified adjustment factor NEAR to carry out compressed encoding as a compression unit independently in the subgraph, but adopts the control method of dynamically-adjusting parameter that the NEAR value is adjusted between subgraph and the subgraph.A concrete execution mode is the division that a subgraph is realized the input remote sensing images with 256 (OK) * 1024 (row) size exactly when the input picture fabric width is 1024.

Two, compressed encoding module

This module is mainly finished the compressed encoding processing of the input subgraph being carried out JPEG-LS, a concrete execution mode is that this compressed encoding module comprises 4 independently compressed encoding units, can carry out the compressed encoding of 4 road images simultaneously handles, with sub-Figure 25 6 (OK) * 1024 (row) size, be divided into 4 bands from height, each band 64 (OK) * 1024 (row), give 4 road compressed encoding units simultaneously with 4 bands and carry out 4 tunnel parallel processings, this module is exported to the Rate Control module with the actual compression code flow CCR of current subgraph (summations of 4 road compressed encodings) 256 (OK) * 1024 (row).

Three, Rate Control module

This module is mainly passed through the deviation of actual compression code flow CCR and targeted compression code flow OCR, the dynamic size of adjusting parameter N EAR value, and feed back to the compressed encoding module, thereby after making different picture materials through the JPEG-LS compression, can both be to carry out the output of compressed bit stream near the code check that requires.The concrete steps of this bit rate control method are for as shown in Figure 6:

(1) the JPEG-LS image with input carries out the subgraph division according to same size, and the subgraph size is r*c, and the pixel that r is capable and c is listed as is promptly arranged, and the subgraph size satisfies

Corresponding one of each subgraph is dynamically adjusted factor NEAR, and each subgraph all is independently processing units; Described dynamic adjustment factor NEAR is the worst error that the picture quality recovery is allowed.

(2) determine dynamically initial value initial_NEAR, minimum M in_NEAR and the maximum Max_NEAR of adjustment factor NEAR according to the JPEG-LS image of target compression ratio and input; After dynamically the minimum M in_NEAR of adjustment factor NEAR and maximum Max_NEAR decide, in the process that the JPEG-LS image of importing compresses, be fixed constant.

Value according to the target compression ratio GR that imports, to the parameter initialize, in the present embodiment, input picture is 10 bit-depths, target compression ratio is that 4 multiplication of voltages contract, and according to Fig. 1, the NEAR value of most of image all concentrates near 4 when 4 multiplication of voltages contract, therefore the initial value of getting NEAR is 4, the maximum of corresponding NEAR value is 9 and 4 multiplication of voltages contract, and minimum value is 2 (having only a width of cloth), for the fluctuation range that makes NEAR not too big, can be based on the situation of most images, the minimum value of NEAR is taken as 3, promptly bigger than normal for the individual image actual compression ratio, but can make the NEAR value of most images adjust level and smooth, in summary, parameter is taken as following value:

Min_NEAR＝3，Max_NEAR＝9，initial_NEAR＝4。

(3) according to the dynamic adjustment factor NEAR that obtains, current subgraph is carried out the JPEG-LS compressed encoding, calculate the cumulative departure amount E (i) of target compression ratio and current compression ratio, that is:

$E\left(i\right)=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}|\mathrm{OCR}-\mathrm{CCR}(i-t)|,n=\{\mathrm{0,1},...,N-1\}$

Here, N is the accumulative total frequency, And N 〉=16, wherein, R*C represent the size of continuously adjustable input picture be R capable * C row, R=4096, the fabric width of C presentation video are the C row, consistent with the fabric width of the JPEG-LS image of input described in the step (1), r*c is the subgraph size.

Input picture fabric width C=1024 in this example, the subgraph size is r=256, c=1024, therefore,

OCR is the targeted compression code flow, and after target compression ratio CR was given, can convert obtained, and reduction formula is Wherein, the input data total amount of subgraph=subgraph line number r* subgraph columns c* subgraph pixel precision/8 bytes, the unit of account here is example with the byte, if the data total amount adopts other unit, can convert on an equal basis), OCR is a constant in entire image compression bit rate adjustment process, E (i) is the corresponding cumulative departure amount of i subgraph (current subgraph), n dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, and n is nonnegative integer, and initial value is 0; CCR (i-t) is the actual compression code flow of i-t subgraph, during as t=0, the actual compression code flow of CCR (i-0)=i subgraph of CCR (i) expression, during t=1, the actual compression code flow of i-1 subgraph of CCR (i-1) expression, be the compressed code flow of a last subgraph of current subgraph (i subgraph), the rest may be inferred for the situation of t=2.

(4) calculate adjustment threshold value t1, the value of t2, that is:

$\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="HSA00000405497200011.png,HSA00000405497200031.png"\; state="\{\{state\}\}">t</figure-callout>}1=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{CCR}(i-t)/\mathrm{CR}*\mathrm{\&alpha;},\mathrm{\&alpha;}=1/8;$

$\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="HSA00000405497200011.png,HSA00000405497200031.png"\; state="\{\{state\}\}">t</figure-callout>}2=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{CCR}(i-t)/\mathrm{CR}*\mathrm{\&beta;},\mathrm{\&beta;}=1/32.$

Wherein, CCR (i-t) is the actual compression code flow of i-t subgraph (256 row image), and CR is a target compression ratio, is 2 multiplication of voltages when contracting as target compression ratio, CR=2, and target compression ratio is 4 multiplication of voltages when contracting, CR=4.N dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, and n is nonnegative integer, and initial value is 0, and span is [0, N-1].

(5) according to the cumulative departure amount E (i) and t1, the t2 that obtain, pass through formula Calculate the updating value of dynamically adjusting factor NEAR;

Wherein, Δ Q is the adjustment step-length, and

NEAR (i) is the value of the dynamic adjustment factor NEAR that gets of i subgraph, NEAR (i-1) is a last subgraph, the i.e. value of the dynamic adjustment factor NEAR that gets of i-1 subgraph, initial_NEAR is an initial value of dynamically adjusting factor NEAR, n dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, N is the accumulative total frequency

And N 〉=16, wherein, R*C represent the size of continuously adjustable input picture be R capable * C row, R=4096, the fabric width of C presentation video are the C row, and be consistent with the fabric width of the JPEG-LS image of input described in the step (1).

Calculate the updating value of adjusting factor NEAR:

if(n＝＝0)

NEAR(i)＝initial_near；

else{

if(E(i)＞t1)

NEAR(i)＝NEAR(i-1)+2；

else?if(E(i)＞t2)

NEAR(i)＝NEAR(i-1)+1；

else?if(E(i)＜-t1)

NEAR(i)＝NEAR(i-1)-2；

else?if(E(i)＜-t2)

NEAR(i)＝NEAR(i-1)-1；

else

NEAR(i)＝NEAR(i-1).

}

Wherein, NEAR (i) represents i the NEAR value that subgraph is got, the value of the NEAR that a last subgraph of i subgraph of NEAR (i-1) expression is got.

(6) the value NEAR (i) of the dynamic adjustment factor NEAR that gets according to i subgraph obtaining carries out clamper by following formula to NEAR (i);

$\mathrm{NEAR}\left(i\right)=\left\{\begin{array}{c}\mathrm{Max}\_\mathrm{NEAR},\mathrm{NEAR}\left(i\right)>\mathrm{Max}\_\mathrm{NEAR}\\ \mathrm{Min}\_\mathrm{NEAR},\mathrm{NEAR}\left(i\right)<\mathrm{Min}\_\mathrm{NEAR}\end{array}\right.,$

Wherein, Min_NEAR is a minimum value of dynamically adjusting factor NEAR, and Max_NEAR is a maximum, in case behind Min_NEAR and the Max_NEAR initialize, be exactly a constant in whole code check adjustment process, immobilizes.

(7) carry out following operation to adjusting frequency n: give n with the n+1 assignment, promptly n=n+1 judges afterwards whether n equals N, if n=N, then make n=0, then forward step (2) to, if n ≠ N, return step (3), circulation is gone down so always, does not have in step (1) till the view data input.

Fig. 4 is a port definition of realizing this module with FPGA, wherein, OCR is the targeted compression code flow, and CCR is the actual compression code flow, the near_start signal is the enable signal of Rate Control module, when this signal was effective, the Rate Control module began to carry out the renewal computing of NEAR value, and next_near is the parameter update value of the next subgraph of output, in the practical application, when no input picture, the near_start enable signal is a disarmed state, and this Rate Control module is not worked.

Four, code stream buffer memory control module

This module is mainly finished the compressed bit stream in the set time section T is cached among the Stand Alone Memory RAM, by the read-write ping-pong operation of two RAM, finishes the continuous output of compressed bit stream.A kind of execution mode is that N continuously adjustable subgraph is cached among the RAM, then the code check output to require; Another kind of execution mode is to do level cache among the RAM with being cached to less than N subgraph, and then carries out the L2 cache of subsequent treatment with the buffer memory that is not less than N subgraph.

The reason of carrying out the code stream buffer memory is usually to need to export constant code rate in actual applications, and after the rate control algorithm convergence, bit rate output will carry out small fluctuation up and down at target bit rate in real time, therefore should not directly export.In conjunction with (be that the continuously adjustable input picture compresses the needed time, R=4096) code stream within carries out after the buffer memory, with the effect that realizes the compressed code flow in the T is averaged, thereby has realized the requirement of constant output code check to a time period T.

Fig. 5 adopts above rate control algorithm, result after the 10 bit remote sensing images of one width of cloth 4096*1024 are compressed, red curve Average CR among the figure represents from beginning the average bit rate output till the current line, blue curve Current CR represents the online bit rate output of current block, as can be seen from the figure, Current CR is having small fluctuating after the algorithmic statement near 4 multiplication of voltages contract curve, but Average CR approaches the 4 multiplication of voltages code check that contracts, and output is stable.

The specific embodiment of the present invention is an example with the bit rate control method of remote sensing images JPEG-LS, but the scope of application of the present invention is not limited to the remote sensing images field.

The unspecified part of the present invention belongs to general knowledge as well known to those skilled in the art.

Claims (2)

1. the bit rate control method of a JPEG-LS image compression is characterized in that step is as follows:

(1) the JPEG-LS image with input carries out the subgraph division according to same size, and the subgraph size is r*c, and the pixel that r is capable and c is listed as is promptly arranged, and the subgraph size satisfies

Corresponding one of each subgraph is dynamically adjusted factor NEAR, and each subgraph all is independently processing units; Described dynamic adjustment factor NEAR is the worst error that the picture quality recovery is allowed;

(2) determine dynamically initial value initial_NEAR, minimum M in_NEAR and the maximum Max_NEAR of adjustment factor NEAR according to the JPEG-LS image of target compression ratio and input; After dynamically the minimum M in_NEAR of adjustment factor NEAR and maximum Max_NEAR decide, in the process that the JPEG-LS image of importing compresses, be fixed constant;

(3) according to the dynamic adjustment factor NEAR that obtains, current subgraph is carried out the JPEG-LS compressed encoding, calculate the cumulative departure amount E (i) of target compression ratio and current compression ratio;

(4) determine dynamically to adjust compare threshold t1 and t2 by following formula:

$t1=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{CCR}(i-t)/\mathrm{CR}*\mathrm{\&alpha;},$ $t2=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}\mathrm{CCR}(i-t)/\mathrm{CR}*\mathrm{\&beta;},$

Wherein, CR is a target compression ratio, and α, β are thresholding coefficient and α=1/8, β=1/32, and CCR (i-t) is the actual compression code flow of i-t subgraph, and when t=0, CCR (i) is the actual compression code flow of i subgraph, and i is a nonnegative integer; N dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, and n is nonnegative integer, and initial value is 0;

(5) according to the t1, the t2 that obtain in cumulative departure amount E (i) that obtains in the step (3) and the step (4), pass through formula

Calculate the updating value of dynamically adjusting factor NEAR;

Wherein, Δ Q is the adjustment step-length, and

NEAR (i) is the value of the dynamic adjustment factor NEAR that gets of i subgraph, NEAR (i-1) is a last subgraph, the i.e. value of the dynamic adjustment factor NEAR that gets of i-1 subgraph, initial_NEAR is an initial value of dynamically adjusting factor NEAR, n dynamically adjusts the adjustment number of times that factor NEAR adjusts numerical value, N is the accumulative total frequency

And N 〉=16, wherein, R*C represent the size of continuously adjustable input picture be R capable * C row, R=4096, the fabric width of C presentation video are the C row, and be consistent with the fabric width of the JPEG-LS image of input described in the step (1);

(6) the value NEAR (i) of the dynamic adjustment factor NEAR that gets according to i the subgraph that obtains in the step (5) carries out clamper by following formula to NEAR (i), enters step (7) afterwards;

$\mathrm{NEAR}\left(i\right)=\left\{\begin{array}{c}\mathrm{Max}\_\mathrm{NEAR},\mathrm{NEAR}\left(i\right)>\mathrm{Max}\_\mathrm{NEAR}\\ \mathrm{Min}\_\mathrm{NEAR},\mathrm{NEAR}\left(i\right)<\mathrm{Min}\_\mathrm{NEAR}\end{array}\right.,$

Wherein, Min_NEAR is a minimum value of dynamically adjusting factor NEAR, and Max_NEAR is a maximum;

(7) carry out following operation to adjusting frequency n: give n with the n+1 assignment, promptly n=n+1 enters step (8) afterwards;

(8) judge whether n equals N,, forward step (2) to, if n ≠ N then forwards step (3) to if n=N then makes n=0.

2. the bit rate control method of a kind of JPEG-LS image compression according to claim 1 is characterized in that: the cumulative departure amount E (i) that calculates target compression ratio and current compression ratio described in the step (3) is undertaken by following formula:

$E\left(i\right)=\underset{t=0}{\overset{n}{\mathrm{\&Sigma;}}}|\mathrm{OCR}-\mathrm{CCR}(i-t)|,n=\{\mathrm{0,1},...,N-1\};$

Wherein, OCR represents the targeted compression code flow, and N is the accumulative total frequency.

Images