CN101126807A - Remote sensing image double nondestruction and near nondestruction code compression method - Google Patents

Abstract

The utility model discloses a doubled-lossless and near-lossless compression coding method on remote sensing image, which is characterized in the process that an original image is divided into a plurality of strips, and the strip coding threshold T is determined according to the width and the height of each strip and the image data precision; all the data of every strip are split into an important bit planar sequence and an unimportant bit planar sequence according to the data split bit of the data; the important bit planar sequence is coded losslessly; the unimportant planar bit stream is output under control by outputting the length of the important bit planar coding stream, the coding threshold and the relationship between the redundant bits so as to try to reach the coding threshold; after being adjusted, the data split bit of the next strip is used to code the next strip; the steps are repeated in sequence to complete the compression coding of all the strips. The utility model has the advantages of low complexity, easy hardware realization, fast speed and high efficiency; thus the method can be used to compress the satellite remote sensing image.

Description

Two times of nothings of remote sensing images undermine nearly lossless compression-encoding method

Technical field:

The present invention relates to the image compression encoding technical field, is a kind of harmless, near-lossless coding method specifically, is used for the compressed encoding of satellite remote sensing images.

Background technology:

In satellite remote sensing is used, carry all kinds of useful load on satellite platform, as visible light camera, multi-spectral imager, infrared camera, interference synthetic aperture radar etc., can obtain different types of remote sensing images.Compare with the general nature image, the correlativity of remote sensing image data a little less than, have information entropy height, characteristics that redundance is low, be unusual difficulty so carry out the real-time coding of high compression ratio.In actual applications, the compression quality of satellite remote sensing images requires with other TV, scenery, medical image very big difference is arranged also, shows:

1). military satellite remote sensing images, especially remote sensing image, some important information spinners will lie in the set of some point, line and point.For the interpretation of military target in the remote sensing images (as tank, aircraft, bridge and airfield runway or the like), the geometric characteristic of main at present dependence military target is discerned.After this just required encoding and decoding to handle, distortion and distortion did not take place in the shape of military target and position.

2). because satellite remote sensing images is taken in hundreds of kilometer high-altitude, be subjected to all multifactor influences such as sun altitude, atmospheric scattering during imaging, cause original image quality relatively poor, signal to noise ratio (S/N ratio) is lower, and contrast is inhomogeneous.This just requires coding/decoding system should be avoided the coding noise that adds up as far as possible, as blocking artifact, convex-concave noise, false figure noise, ribbon noise and " coiling " distortion or the like, causes losing of military target information.

3). compress in order to adapt on the star, algorithm require must easy, quick, shared memory capacity little, be easy to the hardware realization.

Because the original data rate of these remote sensing images is very high, and has information redundancy, therefore before carrying out data storage and transmitting, need remote sensing images are carried out compressed encoding efficiently.In order to keep the quality of original image to greatest extent, under some is used, require these remote sensing images be can't harm or nearly harmless compression.

In recent years, Theory of Image Coding has been obtained a series of achievements that attract people's attention.Shapiro had proposed embedded zerotree image (EZW) method in 1993, Pearlman in 1996 is according to the basic thought of Shapiro zerotree image, a kind of new and implementation method that performance is more excellent has been proposed, promptly ordering (Set Partitioning inHierarchical Trees, encryption algorithm SPIHT) are cut apart in set based on hierarchical tree.The association rate distortion optimized algorithm of Taubman had proposed the EBCOT algorithm in 2000, and this algorithm is adopted by up-to-date international code standard JPEG2000, becomes the core algorithm of JPEG2000 coded system.These high-performance algorithm are applied to picture coding to wavelet transformation, are once new image coding technique revolutions.But these are the compression algorithm of core with the wavelet transformation, because the computation complexity height of wavelet transformation, bit-plane coding complexity height and required storage space are very big, hardware implementation complexity height is difficult to satisfy the requirement of satellite remote sensing images compression applications.Traditional jpeg image compression method compression efficiency is lower, and is unsuitable for harmless or near lossless compress; The anti-error code capacity of jpeg algorithm is not strong simultaneously, can cause serious error code diffusion in the higher satellite channel transmission of the bit error rate, therefore should not adopt yet.

The content of invention

The objective of the invention is to overcome the deficiency of above-mentioned prior art, provide a kind of encoder complexity low, be easy to two times of nothings of hard-wired remote sensing images and undermine nearly lossless compression-encoding method, to realize the requirement of satellite remote sensing images compression applications.

The key problem in technology of realizing the object of the invention is that original image is divided into band, each band absolute coding.In the band coding, view data is divided into significant bits plane and non-significant bits plane, important bit-planes data are carried out lossless coding, whether export non-significant bits panel data according to the output code flow length and the relation decision of coding thresholding, to realize Rate Control.Detailed process is as follows:

1, width, the elevation information according to the input original image is divided into band with image, and determines coding thresholding T according to width, height and the view data precision of each band, promptly

$T=\frac{w\×h\×N}{16}$

Wherein: w is the width of original image

H is the height of band

N is the precision of view data;

2, dark N in position and the data according to input picture split bit bp, and all data in each band are split into significant bits plane sequence sigs _iWith non-significant bits plane sequence errs _i

3, to important bit-planes sequence sigs _iCarry out lossless coding, promptly earlier to important bit-planes sequence sigs _iCarry out difference and mapping and handle non-negative sequence δ of back generation _i, again should non-negative sequence δ _iBe divided into one group by per 16 data, independently carry out entropy coding, complete sequence δ successively encodes _i

4, when the significant bits sequential coding of a band finishes, the output code flow length L is compared with coding thresholding T, be greater than or less than the Different Results of coding thresholding T according to the significant bits plane code stream length L of current band, the data of next band are split a position bp ' adjust, and to the digital coding of this band;

5, successively all bands are finished coding after, export total encoding code stream.

Two times of nothings of above-mentioned remote sensing images undermine nearly lossless compression-encoding method, when wherein the significant bits plane code stream length L of current band meets or exceeds coding thresholding T, then carry out following processing:

(1) if L-T 〉=T/N, the data of then adjusting next band split bit bp ', make bp '=bp+1;

(2) upgrade the redundant bit length R of output code flow length L and thresholding, make the redundant bit length R '=R+L-T after the renewal, enter next band again and encode.

Two times of nothings of above-mentioned remote sensing images undermine nearly lossless compression-encoding method, when wherein the significant bits plane code stream length L of current band is less than total coding thresholding T, then carry out following processing:

(1) if $T-L\≥\frac{w\×h}{8},$ The data of then adjusting next band split bit bp ', make bp '=bp-1, if bp has been 0 and has not processed;

(2) if T-L≤R upgrades current redundant bit length R, make the redundant bit length R '=R+L-T after the renewal, enter next band coding;

(3) if T-L＞R then exceeds the relation of length T-L-R and non-significant bits stream length U by further judgement, determine the output of non-significant bits:

(a) if T-L-R 〉=U exports non-significant bits code stream length U and non-significant bits code stream, upgrade current redundant bit length R, enter next band coding;

(b),, enter next band coding with current redundant bit length R zero clearing if T-L-R＜U exports length information T-L-R and the non-significant bits code stream of part.

Two times of nothings of above-mentioned remote sensing images undermine nearly lossless compression-encoding method, wherein to important bit-planes sequence sigs _iCarrying out difference processing is, successively with each band significant bits plane 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,

$\left\{\begin{array}{c}{\mathrm{ss}}_i\left(0\right)={\mathrm{sigs}}_i\left(0\right)\\ {\mathrm{ss}}_i\left(j\right)={\mathrm{sigs}}_i\left(j\right)-{\mathrm{sigs}}_i(j-1)\end{array}\right.$

Wherein: ss _i(0) and sigs _i(0) is respectively sequence ss _iAnd sigs _iInitial data;

Sigs _i(j) be image band significant bits plane sequence sigs _iIn j the point value;

Ss _i(j) be difference sequence ss _iIn j the point value.

Two times of nothings of above-mentioned remote sensing images undermine nearly lossless compression-encoding method, wherein to important bit-planes sequence sigs _iShining upon processing is with difference sequence ss _iBe mapped as a non-negative sequence δ _i, specifically shine upon formula and be:

Wherein: θ=min (ss _i(j-1), 2 ^N-bp-ss _i(j-1)-1);

N is a data precision, and bp is that the data of present encoding band split the position;

δ _i(j) be sequence δ _iIn j the point value.

Two times of nothings of above-mentioned remote sensing images undermine nearly lossless compression-encoding method, wherein with non-negative sequence δ _iBeing divided into one group by per 16 data, independently carrying out entropy coding, is by adaptive entropy coding device, adopts the division coding method to carry out according to the following procedure:

(1) integrated data n position coded data is carried out data splitting according to division position k, produce two groups of data: the low Bit data of one group of n-k position higher bit data and one group of k position;

(2) the higher bit data are carried out Run-Length Coding;

(3) low Bit data directly is attached to higher bit data code flow back;

(4) use difference to divide the code stream length that the position coding will produce according to dividing the order of position k, calculating successively,, choose the pairing division of the shortest code stream of a code stream length k so the code stream length of generation is compared from 0 to N-bp _Opt, described 16 data are divided coding.

Two times of nothings of above-mentioned remote sensing images undermine nearly lossless compression-encoding method, wherein significant bits plane sequence sigs _iBe that a new sequence is formed separately on former sequence of N-bp higher bit plane, the sequence number of this sequence is followed successively by 0,1 from low to high ... N-bp-1; Wherein non-significant bits plane sequence errs _iBp the low bit-planes that is former sequence is formed another new sequence separately, the sequence number of this sequence is followed successively by 0,1 from low to high ... bp-1.

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 has provided the satellite remote sensing images of the present invention to 14 width of cloth reality, comprise visible images, infrared image and multispectral image carry out 2 multiplication of voltages when contracting the actual compression multiple and recover image Y-PSNR test result.

By table 1 as seen, the Y-PSNR PSNR of recovery image all can reach more than the 60dB when 2 multiplication of voltages contract; Simultaneously as seen, can also realize lossless compress to most of images.Therefore, the present invention is very suitable for harmless, the closely harmless Real Time Compression application of remote sensing images on the satellite.

Table 1 Y-PSNR

Image name (* .raw)	The actual compression multiple	Recover signal noise ratio (snr) of image (dB)
			1	2.14287	Harmless
2	2.13523	Harmless
			3	2.00597	Harmless
4	1.99986	Harmless
			5	1.99932	70.49
6	1.91755	Harmless
			7	1.99932	Harmless
8	1.99695	Harmless
			9	2.06504	Harmless
10	1.99884	74.92
			11	1.99984	98.36
12	1.98332	64.73
			13	1.99984	66.11
14	1.98443	70.44

Description of drawings

Fig. 1 is a coding FB(flow block) of the present invention;

Fig. 2 is that the present invention splits the FB(flow block) that non-significant bits plane code stream was adjusted and exported in the position to data;

Fig. 3 is that the present invention carries out high low level deconsolidation process synoptic diagram to view data;

Fig. 4 is an entropy coding FB(flow block) of the present invention;

Fig. 5 is the structured flowchart of encoding code stream of the present invention.

Embodiment

Present embodiment adopts the width of original input picture and is 512 pixels highly, and swath height is 64 pixels, and the precision of view data is 12 bits, and file header length is appointed as 10 bytes.

See figures.1.and.2, compression encoding process of the present invention is as follows:

The first step is carried out piecemeal to image, determines coding thresholding T.

According to the width of input picture and height image is divided into 8 width and is 512 pixels, highly is the band of 64 pixels, be i.e. 512 * 64=32768 data point.And according to the definite thresholding T that encodes of width, height and the view data precision of each band, according to formula

$T=\frac{w\×h\×N}{16},$ The coding thresholding T of every band is decided to be 512 * 64 * 12/16=24576 bit, this coding thresholding T is the ideal value of band code stream length, for simple image, actual band code stream length can be lower than coding thresholding T, otherwise for complicated image, actual band code stream length also may be higher than coding thresholding T.

Second step, determine that block data splits bit bp, the line data of going forward side by side splits.

Setting initial bp is 2, and dark N in position and data fractionation bit bp according to input picture split into significant bits plane sequence sigs with all data in each band _iWith non-significant bits plane sequence errs _i, this significant bits plane sequence sigs _iA new sequence of forming separately by former sequence of N-bp higher bit plane; This non-significant bits plane sequence errs _iBp another new sequence that low bit-planes is formed separately by former sequence.

For first band, because initial bp=2, its significant bits plane sequence is made up of 10 bit-planes, and non-significant bits plane sequence is made up of 2 bit-planes.If i strip data before the deconsolidation process is designated as s _i, s _iComprise 32768 data points, each data point comprises 12 bit-planes, and promptly the precision of data is 12 bits.In 12 bit-planes, being positioned at uppermost 10 higher bit planes is important bit-planes, with s _iIn 10 higher bit planes of all data points form significant bits plane sequence, be designated as sigs _i, all the other 2 low bit-planes are formed non-significant bits plane sequence, are designated as errs _i, as shown in Figure 3.In Fig. 3, the scale-of-two 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 split presequence and have N bit-planes, its sequence number is followed successively by 0,1 from low to high ... N-1 is through after the high low level deconsolidation process, a new sequence is formed on former sequence of N-bp higher bit plane separately, as significant bits plane sequence sigs _i, this sequence has N-bp bit-planes, its sequence number is followed successively by 0,1 from low to high ... N-bp-1.The bp of former sequence low bit-planes then formed another new sequence separately, as non-significant bits plane sequence errs _i, this sequence has bp bit-planes, its sequence number is followed successively by 0,1 from low to high ... bp-1.

For example, 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.The importance of pressing bit-planes is with s _i(j) split into two parts, a part is preceding 10 bits 0010011101, and corresponding decimal representation is 157, is designated as sigs _i(j), expression significant bits plane sequence sigs _iThe j point; Another part is back 2 bits 01, and corresponding decimal representation is 1, is designated as errs _i(j), represent non-significant bits plane sequence errs _iThe j point.

The 3rd step is to important bit-planes sequence sigs _iCarry out lossless coding.

(1) to current band significant bits plane 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 the value of j point, concrete computing formula is,

$\left\{\begin{array}{c}{\mathrm{ss}}_i\left(0\right)={\mathrm{sigs}}_i\left(0\right)\\ {\mathrm{ss}}_i\left(j\right)={\mathrm{sigs}}_i\left(j\right)-{\mathrm{sigs}}_i(j-1)\end{array}\right.$

Wherein: ss _i(0) and sigs _i(0) is respectively sequence ss _iAnd sigs _iInitial data;

Sigs _i(j) be image band significant bits plane sequence sigs _iIn j the point value;

Ss _i(j) be difference sequence ss _iIn j the point value.

(2) to important bit-planes sequence sigs _iShine upon processing, with difference sequence ss _iBe mapped as a non-negative sequence δ _i, because ss _iIn negative value can appear, to ss _iThe non-negative mapping that tries again generates a new non-negative sequence δ _i, establish δ _i(j) be sequence δ _iIn the value of j point, specifically shine upon formula and be:

Wherein: θ=min (ss _i(j-1), 2 ^N-bp-ss _i(j-1)-1);

N is a data precision, and bp is that the data of present encoding band split the position;

δ _i(j) be sequence δ _iIn j the point value.

(3) with described non-negative sequence δ _iBe divided into one group by per 16 data, independently carry out entropy coding, as shown in Figure 4, its process is as follows:

1) integrated data n position coded data is carried out data splitting according to division position k, produce two groups of data: the low Bit data of one group of n-k position higher bit data and one group of k position;

2) the higher bit data are carried out Run-Length Coding;

3) low Bit data directly is attached to higher bit data code flow back;

4) use difference to divide the code stream length that the position coding will produce according to dividing the order of position k, calculating successively,, choose the pairing division of the shortest code stream of a code stream length k so the code stream length of generation is compared from 0 to N-bp _Opt, described 16 data are divided coding successively.

In the present embodiment, be divided into one group according to per 16 data, every band can divide 2048 groupings, and entropy coding is independently carried out in each grouping successively.In the entropy coding process of each grouping, according to the order of division position k from 0 to N-bp, calculate code stream length of using different divisions position coding to produce successively, choose the shortest division position k of code stream length after relatively _Opt, use the division position to be k to this grouping _OptDivision coding.If the δ before the entropy coding in some groupings _i(j)=0000111010, corresponding decimal number is 72, and it is divided the division coding of a k from 0 to 10, record coding code length respectively.Suppose it is done the division coding of k=3, δ _i(j) at first be split into two data: δ _Ih(j)=0000111, δ _Il(j)=010, δ _Ih(j) corresponding decimal number is 7, so be 00000001 to its code word of doing Run-Length Coding, length is 8, δ _Il(j) code length is 3, obtains δ _i(j) the division coding codeword length of carrying out k=3 is 11.Record k=0,1 ..., 10 o'clock to δ _i(j) code word size of division coding is to 16 δ in this grouping _i(j) do same treatment, and add up respectively and obtain k=0,1 ..., 10 o'clock grouping code stream length, the k value of the code stream correspondence of selection length minimum is designated as k _Opt, then the position is divided in this grouping and is k _OptDivision coding, when output grouping code stream, to export the k of 4 bits earlier _OptValue.

In the 4th step, significant bits plane code stream length L is compared with coding thresholding T, and adjust the data fractionation position bp ' of next band.

When the significant bits sequential coding of a band finishes, the output code flow length L is compared with coding thresholding T, be greater than or less than the Different Results of coding thresholding T according to the significant bits plane code stream length L of current band, the data of next band split a position bp ' adjust, that is:

If L-T 〉=T/N makes bp '=bp+1;

If $T-L\≥\frac{w\×h}{8},$ Then make bp '=bp-1, be 0 as if bp this moment and do not processed, should

Code stream for the bit-planes in non-significant bits plane;

Bp ' is constant under other situations, and for example the first band significant bits plane coding code stream length L is 22639 bits, and T is 24576 bits, and T-L is 1937 bits, Be 4096 bits, so the bp ' of next band is constant.

In the 5th step, export non-significant bits panel data errs according to different situations _i, obtain the encoding code stream of current band.

1. as the code stream L on significant bits plane during more than or equal to coding thresholding T, be L 〉=T, perhaps as L when simultaneously the difference of L and T is no more than current redundant bit length R less than T, be T-L≤R, need not export non-significant bits panel data, only need the non-significant bits stream length information 0 of output 4 bytes to get final product, will upgrade current redundant bit length R simultaneously.

2. as L when simultaneously the difference of L and T surpasses current redundant bit length R less than T, promptly T-L＞R needs the non-significant bits data of output, needs the relation of judgement non-significant bits plane code stream length U and T-L-R this moment:

(1), then exports non-significant bits plane code stream (output length is T-L-R) if U 〉=T-L-R then exports 4 byte length information T-L-R;

(2) if U＜T-L-R then exports 4 byte length information U, then export whole non-significant bits plane code stream, upgrade current redundant bit length R.

For example, in first band, current R is 0, and T-L is 1937 bits, and non-significant bits plane code stream length U is 8192 bits, so need the non-significant bits of output plane code stream to reach band coding thresholding T.Export the length information 1937 of 4 bytes earlier, export the non-significant bits plane code stream of 1937 bits then successively, current redundant bit length R still is 0.

The significant bits plane coding code stream and the 5th that described the 3rd step produces goes on foot non-significant bits plane this two parts sum of code stream of output, constitutes the encoding code stream of current band.

The 6th step was that data split the position with bp ', turned back to second stepping and went into next band coding.

With bp ' serves as to split the position next strip data is split, and carries out data fractionation position, the non-significant bits panel data of output of significant bits plane sequence nondestructive coding, the next band of adjustment successively, obtains the encoding code stream of second band.

The 7th step, carry out second repeatedly and went on foot for the 6th step, obtain the total compression encoding code stream of all bands.

This total compressed encoding code stream by file header and each independently compressed package form, as shown in Figure 5.

With reference to Fig. 5, file header length is fixed, and comprises the information of input picture: wide width, high height, the high H of band and the dark N in pixel position.Each compressed package comprises the encoding code stream of the corresponding band of packet header and this compressed package institute, and its middle wrapping head is 1 byte by length, and the data that are used to preserve current band split position bp composition; The encoding code stream of each band is the significant bits plane coding code stream and the non-significant bits plane code stream sum of this band.This significant bits code stream is made up of each grouping code stream successively, and each code stream that divides into groups comprises 4 bits division position k _OptValue and group coding code stream.

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 at claim protection domain of the present invention.

Claims (8)

1. two times of nothings of remote sensing images undermine nearly lossless compression-encoding method, comprise following process:

(1) width, the elevation information according to the input original image is divided into band with image, and determines coding thresholding T according to width, height and the view data precision of each band, promptly

$T=\frac{w\×h\×N}{16}$

Wherein: w is the width of original image

H is the height of band

N is the precision of view data;

(2) dark N in position and the data according to input picture split bit bp, and all data in each band are split into significant bits plane sequence sigs _iWith non-significant bits plane sequence errs _i

(3) to important bit-planes sequence sigs _iCarry out lossless coding, promptly earlier to important bit-planes sequence sigs _iCarry out difference and mapping and handle non-negative sequence δ of back generation _i, again should non-negative sequence δ _iBe divided into one group by per 16 data, independently carry out entropy coding, complete sequence δ successively encodes _i

(4) when the significant bits sequential coding of a band finishes, the output code flow length L is compared with coding thresholding T, be greater than or less than the Different Results of coding thresholding T according to the significant bits plane code stream length L of current band, the data of next band are split the non-significant bits plane code stream that position bp ' adjusted and exported current band;

(5) be that data split the position with bp ', turn back to (2) and enter next band coding;

(6) carry out (2)～(5) repeatedly, finish the coding of all bands, export total encoding code stream.

2. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, when wherein the significant bits plane code stream length L of current band meets or exceeds coding thresholding T, data to next band split the non-significant bits plane code stream that position bp ' adjusted and exported current band, comprise following process:

1) if L-T 〉=T/N, the data of then adjusting next band split bit bp ', make bp '=bp+1;

2) upgrade current redundant bit length R, make the redundant bit length R '=R+L-T after the renewal;

3) the non-significant bits panel data information of output 4 byte length information 0.

3. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, when wherein the significant bits plane code stream length L of current band is less than total coding thresholding T, data to next band split the non-significant bits plane code stream that position bp ' adjusted and exported current band, comprise following process:

1) if $T-L\≥\frac{w\×h}{8},$ The data of then adjusting next band split bit bp ', make bp '=bp-1, if bp has been 0 and has not adjusted;

2), make the redundant bit length R '=R+L-T after the renewal if T-L≤R upgrades current redundant bit length R; Export the non-significant bits panel data information of 4 byte length information 0;

3) if T-L＞R then exceeds the relation of length T-L-R and non-significant bits code stream length U by further judgement, determine the output of non-significant bits;

(a) if T-L-R 〉=U upgrades current redundant bit length R, export non-significant bits code stream length U and non-significant bits code stream;

(b) if T-L-R＜U with current redundant bit length R zero clearing, exports length information T-L-R and the non-significant bits code stream of part.

4. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, wherein to important bit-planes sequence sgis _iCarrying out difference processing is: successively with each band significant bits plane sequence sgis _iIn the value of each point deduct the value of previous point, obtain a corresponding difference sequence ss _i, specifically computing formula is,

$\left\{\begin{array}{c}{\mathrm{ss}}_i\left(0\right)={\mathrm{sigs}}_i\left(0\right)\\ {\mathrm{ss}}_i\left(j\right)={\mathrm{sigs}}_i\left(j\right)-{\mathrm{sigs}}_i(j-1)\end{array}\right.$

Wherein: ss _i(0) and sigs _i(0) is respectively sequence ss _iAnd sigs _iInitial data;

Sgis _i(j) be image band significant bits plane sequence sigs _iIn j the point value;

Ss _i(j) be difference sequence ss _iIn j the point value.

5. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, wherein to important bit-planes sequence sigs _iShine upon to handle and be: with difference sequence ss _iBe mapped as a non-negative sequence δ _i, specifically shine upon formula and be:

Wherein: θ=min (ss _i(j-1), 2 ^N-bp-ss _i(j-1)-1);

N is a data precision, and bp is that the data of present encoding band split the position;

δ _i(j) be sequence δ _iIn j the point value.

6. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, wherein with non-negative sequence δ _iBeing divided into one group by per 16 data, independently carrying out entropy coding, is by adaptive entropy coding device, adopts the division coding method to carry out according to the following procedure:

(1) integrated data n position coded data is carried out data splitting according to division position k, produce two groups of data: the low Bit data of one group of n-k position higher bit data and one group of k position;

(2) the higher bit data are carried out Run-Length Coding;

(3) low Bit data directly is attached to higher bit data code flow back;

(4) use difference to divide the code stream length that the position coding will produce according to dividing the order of position k, calculating successively,, choose the pairing division of the shortest code stream of a code stream length k so the code stream length of generation is compared from 0 to N-bp _Opt, described 16 data are divided coding.

7. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, wherein significant bits plane sequence sigs _iBe that a new sequence is formed separately on former sequence of N-bp higher bit plane, the sequence number of this sequence is followed successively by 0,1 from low to high ... N-bp-1.

8. two times of nothings of remote sensing images according to claim 1 undermine nearly lossless compression-encoding method, wherein non-significant bits plane sequence errs _iBp the low bit-planes that is former sequence is formed another new sequence separately, the sequence number of this sequence is followed successively by 0,1 from low to high ... bp-1.

Images