CN1622590A - Spatial dimension independent interference multispectral image lossless and near-lossless coding method - Google Patents
Spatial dimension independent interference multispectral image lossless and near-lossless coding method Download PDFInfo
- Publication number
- CN1622590A CN1622590A CN 200410073383 CN200410073383A CN1622590A CN 1622590 A CN1622590 A CN 1622590A CN 200410073383 CN200410073383 CN 200410073383 CN 200410073383 A CN200410073383 A CN 200410073383A CN 1622590 A CN1622590 A CN 1622590A
- Authority
- CN
- China
- Prior art keywords
- bit
- coding
- planes
- significant bits
- sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to image compressing and coding method and solves mainly the problems in complexity and resisting code error. The coding method includes the first determination of image compressing code flow structure, total coding threshold T and the maximum bit plane number max-bp of the whole image; the subsequent circulating shift and higher-lower bit separation pre-treatment of all the input image data; and the successive independent coding of all the important bit plane sequences. During the coding, all the important bit plane sequences are coded until reaching the total coding threshold T, and after coding, the coding code flow is packed and output. In case of not reaching the total coding threshold T, the non-important bit plane sequences are also coded until reaching the total coding threshold T or finishing the all the non-important bit plane sequences. The present invention has the advantages of less complexity, simple hardware and powerful code error resistance.
Description
Technical field:
The present invention relates to the image compression encoding technical field, particularly a kind of space dimension independently can't harm, near-lossless coding method, is used for the compressed encoding of various interference multispectral images.
Background technology:
Along with space remote sensing technology and satellite imagery technology rapid development, a kind of novel remote sensing technology---light spectrum image-forming technology has appearred, caused the revolutionary leap in satellite remote sensing technology field.The light spectrum image-forming technology is melted into a whole optics, spectroscopy, precision optical machinery, electronic technology and computer technology, can obtain the abundant information of the space peacekeeping spectrum dimension of measured target, thereby high using value is all arranged in economic construction and military affairs.It utilize imaging spectrometer on the continuous light spectral coverage to the imaging simultaneously of same target, thereby directly reflect the spectral signature of object being observed.
Imaging spectrometer (Imaging Spectrometer) is a kind of novel air remote sensing equipment, has the function of imager and spectrometer concurrently, can obtain two-dimensional space information and one dimension spectral information simultaneously, and its image-forming principle as shown in Figure 8.In the imaging spectrometer, a branch of incident ray resolves into two bundles and has the light of certain optical path difference and interfere stack through interferometer.Imaging spectrometer adopts row to push away the mode of sweeping and produces interference image since in delegation the optical path difference difference of each point, thereby form multispectral image with several shape interference fringes at imaging plane.When imaging, a point on ground forms a row spectrum, when the inswept ground row luminous point of the aircraft that is loaded with imaging spectrometer, just obtains a complete interference multispectral image, and aircraft constantly pushes away and sweeps, and forms a series of interference multispectral images.According to the principle of Fourier transform spectroscopy, there is the Fourier transform relation between the spectral distribution of interference pattern and light source.
According to image-forming principle, interference multispectral image has following two characteristics:
1. interference multispectral image has very high spatial resolution and spectral resolution, and data volume is very huge.Present ground resolution can reach 8 meters, and spectral resolution can reach 300 spectral coverages, and along with the raising of imaging spectrometer resolution, view data sharply increases, and these all cause huge pressure to real-time Transmission and record.As everyone knows, it is a pair of contradiction that the huge view data rate of required transmission and limited satellite are counted the biography ability, in order to solve this contradiction, just must adopt the image compression encoding technology to satisfy the requirement of view data real-time Transmission.In actual applications, compression process requires to carry out on star, so the complexity of the compaction coding method that adopts can not be too high, otherwise is unfavorable for realizing Real Time Compression.
2. interference multispectral image has correlation between more weak spatial coherence and stronger spectrum, and it is in the degree of correlation difference of level and vertical direction in addition, and the vertical direction degree of correlation is higher than the horizontal direction degree of correlation.Based on this, when interference multispectral image is carried out compressed encoding, should make full use of its spectrum between correlation carry out decorrelation.
In addition, consider the situation of actual channel transmission, the code stream that the compressed coding of interference multispectral image back forms is when channel passes down via satellite, because the instability of channel, the situation of errored bit might occur.And wrong bit information will produce error propagation to the reconstruction that recovers image, has reduced the quality of recovering image.Therefore, when multispectral image is carried out compressed encoding, should guarantee that the compaction coding method that adopts has good error-resilient performance.
At the research of interference multispectral image compaction coding method, all still rare at home and abroad at present.The article Spectral Image Compression (vol.21 that delivers at nineteen ninety J.Optics (Paris) periodical such as French scholar Corinne Mailhes, Paul Vermande, Francis Castanie, No.3,1990, pp.121-132) the middle interference multispectral image compaction coding method that proposes based on the AR model.This method is that each row spectrum employing AR model of interference multispectral image is predicted match, then the difference of match is carried out quantization encoding.If the view data precision is 12 bits, 10 rank AR models are adopted in prediction, and then the 1st point in the initial data do not encoded, direct 12 bits output, and then the value of 9 future positions is quantified as the output of 8 bits respectively, and follow-up match difference all is quantified as the output of 2 bits.This method can more effectively be compressed the interference multispectral image data, and to each row spectrum absolute coding, stronger anti-error code capacity is arranged.But, the process of the setting up more complicated of AR model, the corresponding calculated complexity is very high, if adopt 40 rank AR models, need the multiplication and division computing of 40 * 40 rank matrixes, and, this method can not self adaptation be adjusted the coefficient of AR model, all need to rebulid the AR model when each row spectrum is compressed, calculate corresponding AR model coefficient, so complex calculations are unfavorable for the real-time coding on the star.If reduce the exponent number of AR model, can reduce operand accordingly, but precision of prediction decreases also, cause the quantification effect variation, recover picture quality and reduce.As seen the code efficiency of this method is lower.
In addition, the harmless near Lossless Image Compression standard JPEG-LS of international popular also can be used for the compression to interference multispectral image.This method is a kind of spatial domain compression algorithm based on adaptive prediction, context model and Golomb coding, adopt context model that the data of current input are classified, select the control of different coded systems and respective coding, during the coding current pixel, scan past data earlier, with the pixel that receives previously is condition, by distributing the conditional probability of current pixel, infers current pixel value.When coding, will give the bigger incident of probability than short code assignment, the corresponding less incident branch of probability is equipped with long code word.Because JPEG-LS has adopted context modeling and adaptive coding technology, its code efficiency is quite high, and under the situation of no error code, its compression performance is better than other lossless compression method.But, JPEG-LS does not do any conversion in cataloged procedure, and be based on prediction thought, its decoding decoded data that places one's entire reliance upon, therefore, in transmission compressed bit stream process in case the bit mistake occurs, the mistake of corresponding sampling point is rebuild and will be caused context to be made mistakes, and influence contextual self adaptation adjustment and make mistakes, thereby cause error propagation, cause subsequent decoding incorrect.As seen, JPEG-LS does not have the ability of anti-error code, is not suitable for the compression to interference multispectral image.
The technical scheme of invention:
The purpose of this invention is to provide a kind of space and tie up independently interference multispectral image lossless near-lossless coding method, to overcome existing method algorithm complex height, low, the anti-error code capacity difference of code efficiency and the lower deficiency of recovery picture quality when interference multispectral image is carried out compressed encoding.
The key problem in technology of realizing the object of the invention is that interference multispectral image is divided into independently spectrum row by the space dimension, each row spectrum is independently encoded, and design effective compressed bit stream structure, with realize can't harm, nearly lossless compress, and the error code diffusion is limited in the row scope.
The present invention at first according to the structure of the width of input picture, highly definite image compression code stream, encodes according to the structure of this code stream more as follows:
The first step, according to the width of input picture, highly, view data precision, required compression multiple and compressed package packet header length determines total coding thresholding T, promptly
Wherein: w is the width of original image
H is the height of original image
P is the precision of view data
R is required compression multiple
L is a compressed package packet header length;
Second goes on foot, and is determined the maximal bit number of planes max_bp of entire image by the maximum of input image data;
The 3rd step, according to described maximal bit number of planes max_bp, all data to input picture are carried out preliminary treatment, promptly at first all images data are carried out cyclic shift, along space dimension direction each columns is split into two columns certificates according to the importance by bit-planes then, be significant bits plane sequence and non-significant bits plane sequence, and calculate the effective number of planes valid_bp in the non-significant bits plane;
The 4th step, all significant bits plane sequences are encoded, promptly each column weight is wanted bit-planes sequence sigs
iAfter the difference processing, carry out Run-Length Coding again, and at any time total bit number of present encoding output is compared with total coding thresholding T,, then stop coding and all encoding code streams of packing output if total bit number of present encoding output reaches total coding thresholding T; If after the whole end-of-encodes of all significant bits plane sequences, do not reach total coding thresholding T yet, then carried out for the 5th step;
The 5th step, non-significant bits plane sequence to be encoded, total bit number that coding is exported reaches total coding thresholding T or the whole codings of all non-significant bits plane sequences is finished, and stops whole codings, and all encoding code streams of packing output.
Above-mentioned image compressed code flow point is a series of independently compressed packages, and each compressed package comprises packet header and compressed package code stream, and packet header length is fixed, and compressed package code stream length is relevant with input picture, and this compressed package code stream is further divided into code stream header and coded bit stream.This code stream header comprises the maximal bit number of planes max_bp of original image, the coded-bit plane numerical table bp and the actual coding bit long kilsyth basalt len of all significant bits plane sequences, minimum 3 times of this lengths table len repeated storage, wherein the best of the coded-bit number of planes of maximal bit number of planes max_bp and each significant bits plane sequence represents that figure place is 4 bits, and the best of the code length of each significant bits plane sequence represents that figure place is 11 bits; This coded bit stream comprises significant bits plane coding bit stream and non-significant bits plane coding bit stream, this significant bits plane coding bit stream is made up of multiple row significant bits plane sequence bits stream, each row bit stream is divided into identical many groups of structure again, and each group comprises that k sign bit and k point export bit at the coding on m significant bits plane.
It is with all input image data ring shift left (N-max_bp) positions that above-mentioned cyclic shift is handled, N is the precision of view data, make the pith in the simple image data all be positioned at higher bit-planes, preferentially it is encoded, to satisfy the requirement of lossless compress.
The comparative result that calculates effective number of planes valid_bp in the non-significant bits plane in above-mentioned the 3rd step and be according to the number m on the maximal bit number of planes max_bp of entire image and significant bits plane calculates, that is:
Max_bp＞m, then valid_bp=max_bp-m;
Max_bp≤m, then valid_bp=0.
Difference processing in above-mentioned the 4th step is successively each column weight to be wanted bit-planes sequence sigs
iIn the value of each point deduct the value of previous point, obtain a corresponding difference sequence ss
i, specifically computing formula is,
Wherein: sigs
i(j) be important bit-planes sequence sigs
iIn j the point value
Ss
i(j) be difference sequence ss
iIn j the point value
Sign (j) is ss
i(j) corresponding sign bit.
Run-Length Coding process in above-mentioned the 4th step is:
(1) statistical difference sub-sequence ss
iThe maximal bit number of planes, as the coded-bit number of planes of these row;
(2) with difference sequence ss
iBe slit into t group coding successively by every k point one component, require to satisfy k * t=h, h is the height of image;
(3) to each group, export earlier the sign bit of k point successively, press k bit on each bit-planes of bit-planes sequential encoding from high to low again, if this k bit all is 0, then export one 0 bit, otherwise one 1 bit of output is exported this k bit, then successively up to having compiled this difference sequence ss earlier
iAll points or the total coding of arrival thresholding T finish coding;
(4) coded-bit number of planes of these row and the actual coding length of Ben Lie are write the code stream header.
Above-mentioned the 5th the step in to non-significant bits plane sequence errs
iCoding be, to non-significant bits plane, scan successively by bit-planes order from high to low, on current bit-planes,, finish all codings if valid_bp equals 0; If valid_bp is greater than 0, press leu time each non-significant bits plane sequence of output all bits on current bit-planes, and valid_bp subtracted the cataloged procedure that enters next bit-planes after 1, and be 0 or reach total coding thresholding T up to valid_bp, finish all codings.
The present invention is owing to matrix operation that does not have complexity or context prediction, thereby with respect to prior art, complexity is low, and hardware is realized simple, and coding rate and efficient are all than higher, and recovery picture quality satisfies application requirements.Table 1 provided the present invention to the interference multispectral image of 10 width of cloth actual measurement carry out 2 multiplication of voltages when contracting the actual compression multiple and recover the image Y-PSNR.By table 1 as seen, the Y-PSNR (PSNR) that recovers image when 2 multiplication of voltages contract all can reach about 60dB, is enough to satisfy application requirements, as seen can also realize lossless compress to some simple image simultaneously.In addition, because the present invention has taked the space to tie up independently coding method, and to all significant bits plane sequence sigs
iCoded-bit lengths table len repeated storage method, make when code stream produces bit error code in the Channel Transmission process, no matter this error code is arranged in the header or the compression bit stream of code stream, error code diffusion during image reconstruction all is limited within the row scope, improved anti-error code capacity greatly, shown in Fig. 6,7.The present invention is very suitable for harmless, the closely harmless Real Time Compression application of interference multispectral image on the star.
Table 1 Y-PSNR
| Image name ( *.raw) | The actual compression multiple | Recover signal noise ratio (snr) of image (dB) |
| ????00 | ????2.0024 | ????60.70 |
| ????01 | ????2.0024 | ????58.32 |
| ????02 | ????2.0024 | ????59.29 |
| ????03 | ????2.2800 | Harmless |
| ????04 | ????2.0024 | ????59.79 |
| ????05 | ????2.0024 | ????59.98 |
| ????06 | ????2.0024 | ????54.78 |
| ????07 | ????2.0024 | ????60.97 |
| ????08 | ????2.0024 | ????57.82 |
| ????09 | ????2.0024 | ????58.84 |
| ????10 | ????2.2518 | Harmless |
Description of drawings
Fig. 1 is a coding FB(flow block) of the present invention
Fig. 2 hardware system schematic diagram of the present invention
Fig. 3 is that the present invention carries out high low level to a row spectrum and splits pretreated schematic diagram
Fig. 4 is the structured flowchart of encoding code stream of the present invention
Fig. 5 (a) is the coded scanning sequential schematic that any column weight of the present invention is wanted the bit-planes sequence
Fig. 5 (b) is the coded scanning sequential schematic of all significant bits plane sequences of the present invention
Fig. 5 (c) is the coded scanning sequential schematic of each non-significant bits plane sequence of the present invention on a bit-planes
Fig. 5 (d) is the coded scanning sequential schematic on the non-significant bits of the present invention plane
Fig. 6 is that the error rate is 10
-5The time, the present invention and JPEG-LS are to the comparative graph of the recovery image Y-PSNR (PSNR) of 60 width of cloth interference multispectral images compression
Fig. 7 is that the error rate is 10
-5The time, the present invention and JPEG-LS are to the recovery image comparison figure of same width of cloth interference multispectral image compression, and wherein Fig. 7 (a) is the width of cloth input picture in 60 width of cloth interference multispectral images, and Fig. 7 (b) is 10 for the error rate
-5The time, the recovery image that adopts the inventive method to obtain, Fig. 7 (c) is 10 for the error rate
-5The time, the recovery image that adopts the JPEG-LS method to obtain
Fig. 8 is the imaging spectrometer imaging schematic diagram
Embodiment
Present embodiment adopts the width of input picture and highly is 128 pixels, the precision of view data is 12 bits, compression multiple is 2 times, packet header length is appointed as 120 bits, the thresholding T that then always encodes should be 128 * 128 * 12/2-120=98184 bit, should total coding thresholding T also be the upper limit of compressed package code stream length, but for simple image, actual compressed package code stream length can be lower than total coding thresholding T.12 bit-planes of this view data are divided according to importance, and 8 wherein higher bit-planes are important bit-planes, and 4 lower bit-planes are non-significant bits plane.And important bit-planes sequence is carried out sequence after the difference processing be slit into 16 groups by per 8 point one components and encode successively.
Set encoding code stream by each independently compressed package forms, as shown in Figure 4, the compress coding data of compressed package correspondence one width of cloth input picture.Each compressed package comprises packet header and compressed package code stream, its middle wrapping head length is fixed as 120 bits, the compressed package code stream is divided into two parts again, be code stream header and coded bit stream, wherein the code stream header comprises the actual coding bit long kilsyth basalt len of the coded-bit plane numerical table bp of maximal bit number of planes max_bp, all significant bits plane sequences of original image and repeated storage 3 times, and coded bit stream then comprises significant bits plane coding bit stream and non-significant bits plane coding bit stream.Because picture traverse is 128, so total number of significant bits plane sequence is 128.Because view data has 12 bit-planes at most, when adopting binary representation, be enough to represent with 4 bits, thereby the coded-bit plane number average of the maximal bit number of planes max_bp in the code stream header and each significant bits plane sequence is represented with 4 bits, the code length of each significant bits plane sequence all is no more than 16384 bits, so adopting 11 bits represents, thereby the total length of coded-bit plane numerical table bp is 512 bits, the total length of coded-bit lengths table len is 4224 bits, so the total length of code stream header is 4740 bits.
Present embodiment adopts XILINX ISE 5.1 integrated developing software and VHDL language to coding method shown in Figure 1, on the XCV600e-6BG432 of XILINX company field programmable gate array (FPGA) chip, realize, promptly form converter, preprocessor and bit-plane coding device hardware circuit, as shown in Figure 2 by this chip.Wherein, preprocessor comprises cyclic shift processor and high low level deconsolidation process device, and the bit-plane coding device comprises significant bits plane sequential coding device and non-significant bits plane sequential coding device.Whole hardware realizes that the process of coding is as follows:
1. the serial data stream of piece image is realized string and conversion through converter, form 128 * 128 parallel image data, each view data is represented with 12 bits, and add up the maximum of these 16384 data, according to the definite maximal bit number of planes max_bp of maximum, promptly equal with the required number of bits of this maximum of binary representation with entire image.
2. the view data after all conversion is sent into preprocessor and carry out preliminary treatment, at first finish cyclic shift to view data by the cyclic shift processor, promptly according to the maximal bit number of planes max_bp that in converter, adds up the entire image that obtains, with all images datacycle move to left (12-max_bp) position, make the significant bits part in the simple image data all be promoted to higher bit-planes, preferentially it is encoded, to satisfy the requirement of lossless compress, send into high low level deconsolidation process device again; The data of high low level deconsolidation process device after to cyclic shift, along space dimension direction, each columns is split into two columns certificates according to the importance by bit-planes, be significant bits plane sequence and non-significant bits plane sequence, wherein significant bits plane sequence is made up of 8 bit-planes, non-significant bits plane sequence is made up of 4 bit-planes, calculate effective number of planes valid_bp in the non-significant bits plane according to the comparative result of the number 8 on the maximal bit number of planes max_bp of entire image and significant bits plane then, promptly
Max_bp＞8, then valid_bp=max_bp-8;
Max_bp≤8, then valid_bp=0;
If the i columns before the deconsolidation process is according to being designated as s
i, s
iComprise 128 data points, each data point comprises 12 bit-planes, and in 12 bit-planes, being positioned at uppermost high 8 bit-planes is important bit-planes, with s
iIn high 8 bit-planes of all data points form significant bits plane sequence, be designated as sigs
i, all the other low 4 bit-planes are formed non-significant bits plane sequence, are designated as errs
i, as shown in Figure 3.In Fig. 3, the binary representation of the value of a point in the corresponding sequence is shown in each tabulation of 3 frames, and each row is represented a bit-planes, and high-rise more bit-planes is important more, and along with the reduction of bit-planes, importance is successively decreased thereupon.Because the view data precision is N, thereby the fractionation presequence has N bit-planes, its sequence number is followed successively by 0,1 from low to high ... N-1, wherein, low N-m bit-planes is non-significant bits plane, its sequence number is from 0 to N-m-1, and a corresponding high m bit-planes is important bit-planes, and its sequence number is from N-m to N-1.Through high low level deconsolidation process, high m the bit-planes that splits presequence formed a new sequence separately, be called significant bits plane sequence, this sequence has m bit-planes, its sequence number is followed successively by 0,1 from low to high ... m-1, low N-m the bit-planes that splits presequence then formed another new sequence separately, is called non-significant bits plane sequence, this sequence has N-m bit-planes, its sequence number is followed successively by 0,1 from low to high ... N-m-1.If split presequence s
iIn j point s
i(j) value is 629, and corresponding binary form is shown 001001110101, and wherein leftmost 0 bit is positioned at higher bit plane, i.e. the 11st bit-planes, and rightmost 1 bit is positioned at the lowest bit plane, i.e. the 0th bit-planes.s
i(j) preceding 8 bit representation significant bits planes, the then non-significant bits of 4 bit representations plane in.The importance of pressing bit-planes is with s
i(j) split into two parts, a part is preceding 8 bits 00100111, and corresponding decimal representation is 39, is designated as sigs
i(j), expression significant bits plane sequence sigs
iThe j point; Another part is back 4 bits 0101, and corresponding decimal representation is 5, is designated as errs
i(j), represent non-significant bits plane sequence errs
iThe j point.
3. obtain the two-way dateout through preprocessor, a circuit-switched data is important bit-planes sequence, sends into significant bits plane sequential coding device, and non-significant bits plane sequential coding device is sent in its coding output; Another circuit-switched data is a non-significant bits plane sequence, directly sends into non-significant bits plane sequential coding device and encodes, and all encoding code streams of packing output.
In significant bits plane sequential coding device, earlier the i column weight of input is wanted bit-planes sequence sigs
iCarry out difference processing, promptly successively with this significant bits plane sequence sigs
iIn the value of each point deduct the value of previous point, thereby obtain a corresponding difference sequence ss
i, establish sigs
i(j) be important bit-planes sequence sigs
iIn j the point value, ss
i(j) be difference sequence ss
iIn j the point value, sign (j) is ss
i(j) if corresponding sign bit is sigs
i(j)＜sigs
i(j-1), the difference of promptly subtracting each other is a negative, then corresponding sign bit sign (j) is put 1, and to difference negate, i.e. ss
i(j)=0-(sigs
i(j)-sigs
i(j-1)); If sigs
i(j)＞sigs
i(j-1), promptly difference is a positive number, then directly sign bit sign (j) is put 0, and specifically computing formula is,
Judge this difference sequence ss then
iWhether all be 0, if all be 0, then the coded-bit number of planes and the code length of these row, with its code stream header and skipping difference sequence ss of writing direct if being 0
iCoding; If all be not 0, difference sequence ss then
iPress bit-planes order from high to low, successively m bit-planes carried out Run-Length Coding.When carrying out Run-Length Coding, the coded-bit number of planes of these row of statistics is difference sequence ss earlier
iThe maximal bit number of planes.Difference sequence ss in the present embodiment
iBeing slit into 16 groups by per 8 point one components encodes successively, to each group, export the wherein sign bit of 8 points earlier successively, again by encode successively 8 bits on each bit-planes of bit-planes order from high to low, if this 8 bit all is 0, then coding is exported one 0 bit, otherwise exports one 1 bit earlier, then export this 8 bit successively, up to having compiled this difference sequence ss
iAll points or the total coding of arrival thresholding T finish coding, and the coded-bit number of planes and the code length that will be listed as at last write the code stream header.
4. after if all 128 column weights are wanted the whole end-of-encodes of bit-planes sequence, do not reach total coding thresholding T yet, then the non-significant bits of 128 row plane sequence is encoded by non-significant bits plane sequential coding device, scan low 4 non-significant bits planes successively by bit-planes order from high to low during coding, when scanning some bit-planes, judge that at first whether effective number of planes valid_bp in the non-significant bits plane is greater than 0, if valid_bp equals 0, finish all codings, and all encoding code streams of packing output; If valid_bp＞0, promptly the bit value of 128 row non-significant bits plane sequences on current bit-planes is effectively, press leu time each non-significant bits plane sequence of output all bits on current bit-planes this moment, effective then non-significant bits number of planes valid_bp subtracts 1, enter the cataloged procedure of next bit-planes, be 0 or reach total coding thresholding T up to valid_bp, finish all codings, and all encoding code streams of packing output.
With reference to Fig. 5 (a), establishing the i column weight, to want the bit-planes sequence be sigs
i, the horizontal direction of square frame is represented each point of significant bits plane sequence among the figure, its sequence number from left to right is followed successively by 0,1 ... h-1, vertical direction is then represented each bit-planes of significant bits plane sequence, be followed successively by 0,1 from the bottom up ... m-1.With sequence sigs
iH point be divided into t group, sequence number is from 0 to t-1, each group comprises k point, grouping of each little box indicating among the figure.To important bit-planes sequence sigs
iCoded scanning order as follows: at first encode the 0th group, press bit-planes importance order from high to low, in this group of encoding successively 0 to this k of k-1 o'clock m-1, m-2 ..., the bit value on 0 bit-planes, behind the m of this group of the having encoded bit-planes, with the 1st group of same sequential encoding, the 2nd group,,, finish the i column weight and want bit-planes sequence sigs up to all t the groups of having encoded
iCoding.
Square frame among Fig. 5 (b) and Fig. 5 (c) is all represented the entire image size, and width is w, highly is h, and horizontal direction is a space dimension direction, and vertical direction is a spectrum dimension direction.Difference is that each point comprises m significant bits plane among Fig. 5 (b), and each point only comprises a non-significant bits plane among Fig. 5 (c).
With reference to Fig. 5 (b), want the bit-planes sequence along the total w column weight of space dimension direction, its sequence number is followed successively by 0,1 ... w-1, wherein each column weight wants the bit-planes sequence to include h point along spectrum dimension direction, its sequence number is respectively 0,1 ... h-1.Each significant bits plane sequence absolute coding, its coded scanning are that the 0th column weight of at first encoding is wanted the bit-planes sequence in proper order, the respective coding order is with reference to the order among Fig. 5 (a), the 1st row, the 2nd of encoding then are listed as ..., all w column weights are wanted the bit-planes sequence up to having encoded.
With reference to Fig. 5 (c), total w is listed as non-significant bits plane sequence along space dimension direction, its sequence number is followed successively by 0,1 ... w-1, wherein each is listed as non-significant bits plane sequence and includes h point along spectrum dimension direction, its sequence number is respectively 0,1 ... h-1.On arbitrary non-significant bits plane, each non-significant bits plane sequence scans output successively by order from left to right.
With reference to Fig. 5 (d), the vertical direction of square frame is represented each bit-planes of non-significant bits plane sequence among the figure, its sequence number is followed successively by 0,1 from the bottom up ... N-m-1, during to the non-significant bits flat scanning of this N-m, scan the N-m-1 bit-planes earlier, scanning sequency on this bit-planes scans next bit-planes then with reference to the order among Fig. 5 (c), up to the 0th bit-planes.
Fig. 6,7 is the anti-error code effect comparison diagram of the present invention and JPEG-LS method, with reference to Fig. 6,10
-5Under the error rate situation, adopt the inventive method and JPEG-LS method to compress respectively to 60 width of cloth interference multispectral images, by to the Y-PSNR (PSNR) that recovers image as seen, having under the situation of error code, the Y-PSNR (PSNR) of the recovery image that employing the inventive method obtains, though in various degree decline is arranged than the Y-PSNR (PSNR) under the no error code situation, but all remain on more than the 30dB, and the image of half nearly arranged, its Y-PSNR (PSNR) does not almost have anything to reduce, and the Y-PSNR of all images (PSNR) all is higher than the Y-PSNR (PSNR) of the recovery image that employing JPEG-LS method obtains under same bit error rate far away.
With reference to Fig. 7, Fig. 7 (a) is the width of cloth input picture in 60 width of cloth interference multispectral images, and Fig. 7 (b) is 10 for the error rate
-5The time, the recovery image that adopts the inventive method to obtain, Fig. 7 (c) is 10 for the error rate
-5The time, the recovery image that adopts the JPEG-LS method to obtain.Obviously, when code stream produces error code, the recovery image that adopts the inventive method to obtain does not almost produce the error code diffusion, the mistake that its decoding produces only is confined within the row scope, do not influence the subsequent decoding process, those row other all row afterwards that therefore produce error code all can correctly recover.The then distortion fully of recovery image of adopting the JPEG-LS method to obtain, the error code diffusion is very serious.This shows that the inventive method can reduce the error code diffusion greatly, improve and recover picture quality.
For those skilled in the art; after having understood content of the present invention and method; can be under the situation that does not deviate from the principle and scope of the present invention; the method according to this invention is carried out various corrections and the change on form and the details, but these are based on correction of the present invention with change still within claim protection range of the present invention.
Claims (9)
1. independently multispectral image lossless, nearly lossless compression-encoding method are tieed up in a space, at first according to the structure of the width of input picture, highly definite image compression code stream, encode as follows according to the structure of this code stream again:
The first step, according to the width of input picture, highly, view data precision, required compression multiple and compressed package packet header length determines total coding thresholding T, promptly
Wherein: w is the width of original image
H is the height of original image
P is the precision of view data
R is required compression multiple
L is a compressed package packet header length;
Second goes on foot, and is determined the maximal bit number of planes max_bp of entire image by the maximum of input image data;
The 3rd step, according to described maximal bit number of planes max_bp, all data to input picture are carried out preliminary treatment, promptly at first all images data are carried out cyclic shift, along space dimension direction each columns is split into two columns certificates according to the importance by bit-planes then, be significant bits plane sequence and non-significant bits plane sequence, and calculate the effective number of planes valid_bp in the non-significant bits plane;
The 4th step, all significant bits plane sequences are encoded, promptly each column weight is wanted bit-planes sequence sigs
iAfter the difference processing, carry out Run-Length Coding again, and at any time total bit number of present encoding output is compared with total coding thresholding T,, then stop coding and all encoding code streams of packing output if total bit number of present encoding output reaches total coding thresholding T; If after the whole end-of-encodes of all significant bits plane sequences, do not reach total coding thresholding T yet, then carried out for the 5th step;
The 5th step, non-significant bits plane sequence to be encoded, total bit number that coding is exported reaches total coding thresholding T or the whole codings of all non-significant bits plane sequences is finished, and stops whole codings, and all encoding code streams of packing output.
2. coding method according to claim 1, it is characterized in that described image compressed code flow point is a series of independently compressed packages, each compressed package comprises packet header and compressed package code stream, packet header length is fixed, compressed package code stream length is relevant with input picture, and this compressed package code stream is further divided into code stream header and coded bit stream.
3. coding method according to claim 2, it is characterized in that described code stream header comprises the maximal bit number of planes max_bp of original image, the coded-bit plane numerical table bp and the actual coding bit long kilsyth basalt len of all significant bits plane sequences, minimum 3 times of this lengths table len repeated storage, wherein the best of the coded-bit number of planes of maximal bit number of planes max_bp and each significant bits plane sequence represents that figure place is 4 bits, and the best of the code length of each significant bits plane sequence represents that figure place is 11 bits.
4. coding method according to claim 2, it is characterized in that described coded bit stream comprises significant bits plane coding bit stream and non-significant bits plane coding bit stream, this significant bits plane coding bit stream is made up of multiple row significant bits plane sequence bits stream, and each row bit stream is divided into identical many groups of structure again.
5. coding method according to claim 1, it is characterized in that it is with all input image data ring shift left (N-max_bp) positions that described cyclic shift is handled, N is the precision of view data, make the pith in the simple image data all be positioned at higher bit-planes, preferentially it is encoded, to satisfy the requirement of lossless compress.
6. coding method according to claim 1, the comparative result that it is characterized in that calculating in the 3rd step effective number of planes valid_bp in the non-significant bits plane and be according to the number m on the maximal bit number of planes max_bp of entire image and significant bits plane calculates, that is:
Max_bp＞m, then valid_bp=max_bp-m;
Max_bp≤m, then valid_bp=0.
7. coding method according to claim 1 is characterized in that the difference processing in the 4th step is successively each column weight to be wanted bit-planes sequence sigs
iIn the value of each point deduct the value of previous point, obtain a corresponding difference sequence ss
i, specifically computing formula is,
Wherein: sigs
i(j) be important bit-planes sequence sigs
iIn j the point value
Ss
i(j) be difference sequence ss
iIn j the point value
Sign (j) is ss
i(j) corresponding sign bit.
8. coding method according to claim 1 is characterized in that the Run-Length Coding process in the 4th step is:
(1) statistical difference sub-sequence ss
iThe maximal bit number of planes, as the coded-bit number of planes of these row;
(2) with difference sequence ss
iBe slit into t group coding successively by every k point one component, require to satisfy k * t=h, h is the height of image;
(3) to each group, export earlier the sign bit of k point successively, press k bit on each bit-planes of bit-planes sequential encoding from high to low again, if this k bit all is 0, then export one 0 bit, otherwise one 1 bit of output is exported this k bit, then successively up to having compiled this difference sequence ss earlier
iAll points or the total coding of arrival thresholding T finish coding;
(4) coded-bit number of planes of these row and the actual coding length of Ben Lie are write the code stream header.
9. coding method according to claim 1 is characterized in that in the 5th step non-significant bits plane sequence errs
iCoding be, to non-significant bits plane, scan successively by bit-planes order from high to low, on current bit-planes,, finish all codings if valid_bp equals 0; If valid_bp is greater than 0, press leu time each non-significant bits plane sequence of output all bits on current bit-planes, and valid_bp subtracted the cataloged procedure that enters next bit-planes after 1, and be 0 or reach total coding thresholding T up to valid_bp, finish all codings.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100733831A CN100361500C (en) | 2004-12-15 | 2004-12-15 | Spatial dimension independent interference multispectral image lossless and near-lossless coding method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100733831A CN100361500C (en) | 2004-12-15 | 2004-12-15 | Spatial dimension independent interference multispectral image lossless and near-lossless coding method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1622590A true CN1622590A (en) | 2005-06-01 |
| CN100361500C CN100361500C (en) | 2008-01-09 |
Family
ID=34765312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2004100733831A Expired - Fee Related CN100361500C (en) | 2004-12-15 | 2004-12-15 | Spatial dimension independent interference multispectral image lossless and near-lossless coding method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100361500C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107005254A (en) * | 2015-06-12 | 2017-08-01 | 华为技术有限公司 | The methods, devices and systems of channel decoding |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6549666B1 (en) * | 1994-09-21 | 2003-04-15 | Ricoh Company, Ltd | Reversible embedded wavelet system implementation |
| US6816284B1 (en) * | 2000-04-07 | 2004-11-09 | Color Aix Perts Gmbh | Multispectral color reproduction system with nonlinear coding |
-
2004
- 2004-12-15 CN CNB2004100733831A patent/CN100361500C/en not_active Expired - Fee Related
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107005254A (en) * | 2015-06-12 | 2017-08-01 | 华为技术有限公司 | The methods, devices and systems of channel decoding |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100361500C (en) | 2008-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2446615C2 (en) | Adaptive coefficient scanning for video coding | |
| CN1223959C (en) | Appts. and method for macroblock based rate control in coding system | |
| CN1108062C (en) | Method and arrangement of coding and decoding video signal by movement estimation based on characteristic point | |
| US9332258B2 (en) | Method and device for display stream compression | |
| CN1767655A (en) | Multi view point video image parallax difference estimating method | |
| CN1260971C (en) | Image coding appliance and image coding method | |
| CN103650496A (en) | Pixel-based intra prediction for coding in HEVC | |
| EP2773122A1 (en) | Method and device for display stream compression | |
| CN101039430A (en) | Method for scanning quickly residual matrix in video coding | |
| CN1669235A (en) | Method and apparatus for transcoding between hybrid video CODEC bitstreams | |
| CN1589023A (en) | Coding and decoding method and device for multiple coded list lengthening based on context | |
| CN107105208A (en) | A kind of lossless coding and coding/decoding method of Bayer images | |
| CN101126807A (en) | Remote sensing image double nondestruction and near nondestruction code compression method | |
| CN1149853C (en) | Coding and decoding device with movement compensation | |
| CN102014283A (en) | First-order difference prefix notation coding method for lossless compression of image data | |
| CN1405735A (en) | Colour-picture damage-free compression method based on perceptron | |
| CN111669588B (en) | Ultra-high definition video compression coding and decoding method with ultra-low time delay | |
| CN1665299A (en) | Method for designing architecture of scalable video coder decoder | |
| CN1159922C (en) | Method for encoding binary shape signal | |
| CN1777286A (en) | Image processing devices and methods | |
| US9948928B2 (en) | Method and apparatus for encoding an image | |
| RU2419246C1 (en) | Method to compress and recover fixed halftone video images | |
| CN109672891B (en) | Lossless secondary compression method of JPEG image | |
| CN1622590A (en) | Spatial dimension independent interference multispectral image lossless and near-lossless coding method | |
| CN1139353A (en) | Method and apparatus for encoding image signal having object by using extension-interpolation technique |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080109 Termination date: 20141215 |
|
| EXPY | Termination of patent right or utility model |