CN107124612A - The method for compressing high spectrum image perceived based on distributed compression - Google Patents
The method for compressing high spectrum image perceived based on distributed compression Download PDFInfo
- Publication number
- CN107124612A CN107124612A CN201710280701.9A CN201710280701A CN107124612A CN 107124612 A CN107124612 A CN 107124612A CN 201710280701 A CN201710280701 A CN 201710280701A CN 107124612 A CN107124612 A CN 107124612A
- Authority
- CN
- China
- Prior art keywords
- wave band
- area
- band
- image
- interest
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/157—Assigned coding mode, i.e. the coding mode being predefined or preselected to be further used for selection of another element or parameter
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/167—Position within a video image, e.g. region of interest [ROI]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention provides a kind of method for compressing high spectrum image perceived based on distributed compression, and method includes：Based on the information entropy of the correlation of each wave band spectrum and each wave band spectrum in high spectrum image, it is divided into a low correlation band group and multiple high correlation band groups, each high correlation band group includes：One refers to wave band and multiple non-reference wave bands；From with reference to determination area-of-interest and background area in wave band；For each non-reference wave band, refer to area-of-interest, background area in wave band with the group and carry out difference processing respectively, obtain each self-corresponding residual image；Area-of-interest in each high correlation band group, background area and the corresponding residual image of each non-reference wave band, the image of low correlation wave band are compressed successively；The code stream of all compressed encodings is sent.The above method carries out different distributed compressions to different wave bands and different regions and handled, the important information for the high spectrum image that adequately protected, while improving the compression ratio of high spectrum image.
Description
Technical field
The present invention relates to Compression of hyperspectral images technology, particularly a kind of high spectrum image perceived based on distributed compression
Compression method.
Background technology
High-spectrum seem while the set of multiple band images composition comprising spatial information and spectral information, by
Many fields are applied to, such as agricultural, military, geological prospecting and environmental monitoring.However as spatial resolution and spectrum point
The continuous improvement of resolution, brings the data of magnanimity.The data of these magnanimity to high spectrum image storage, transmit and apply band
Carry out huge challenge, therefore how efficiently to realize high-spectral data compression just into urgent problem to be solved.
However, in numerous Hyperspectral image compression algorithms of proposition, most of algorithm is in high spectrum image
Most of wave bands and area of space all take identical to handle, and this may cause some important informations in spatially and spectrally dimension
Loss.In actual applications, high spectrum image professional target interested is often positioned in HFS and is very little
All wave bands and region are taken identical to handle by one piece of region, it is possible to can cause the loss of information of interest.
In addition, the information of interest of high spectrum image includes wave band interested and area-of-interest.And in the prior art can
Enough support the 3D-SPIHT compression algorithms of high spectrum image information of interest protection, the method for being fundamentally based on conversion, threshold value
Constantly update and make it that its amount of calculation and complexity are higher, also make it that its reconstructed image blocking effect is tighter the reason for quantization
Weight, in addition, the 3D-SPIHT compression algorithms could not make full use of high spectrum image very strong spectrum correlation, removes redundancy not thorough enough
Bottom.
The content of the invention
For defect of the prior art, the present invention provides a kind of Compression of hyperspectral images perceived based on distributed compression
Method, area-of-interest and wave band interested to high spectrum image carry out selective distribution formula compression processing, improve image
Flexibility in compression process.
In a first aspect, the present invention provides a kind of method for compressing high spectrum image perceived based on distributed compression, including：
Step 01：For pending high spectrum image, based on each wave band correlation in high spectrum image and each wave band
All wave bands of high spectrum image are divided into a low correlation band group and multiple high correlation band groups by information entropy,
Each high correlation band group includes：The reference wave band and multiple non-reference wave bands of one high entropy；
Step 02：For each high correlation band group, according to default area-of-interest selection strategy, from referring to wave band
Middle determination area-of-interest, will be used as background area with reference to the region in wave band in addition to area-of-interest；
Step 03：For each non-reference wave band of high correlation band group, region of interest in wave band is referred to the group
Domain, background area carry out difference processing respectively, and acquisition corresponds respectively to the area-of-interest and the residual image of background area；
Step 04：Using area-of-interest in each high correlation band group of the first coded system successively compressed encoding and many
The individual residual image associated with area-of-interest；
Using background area and multiple and background in each high correlation band group of the second coded system successively compressed encoding
The residual image of region association；
Image after the sparse transformation of low correlation band group is encoded using the 3rd coded system successively independent compression；
Step 05：The code stream of all compressed encodings is sent.
Alternatively, step 01 includes：
Obtain the information entropy of all wave bands in the high spectrum image；It will be greater than the first corresponding band group of default entropy
Into first set s1；
Obtain the correlation coefficient r between all adjacent bands in the high spectrum image；It is determined that more than the first parameter preset
Those wave bands are constituted second set s by corresponding two wave bands of each coefficient correlation of value2；
By first set s1With second set s2Common factor be used as the 3rd set s3；By first set s1It is middle to remove the 3rd collection
Close s3Element be used as a low correlation band group；
The wave band that low correlation wave band is removed in all wave bands of high spectrum image is divided into multiple high correlation wave bands
Group, it is ensured that there is the 3rd set s in each high correlation band group3In at least one element；
The wave band that low correlation wave band is removed in all wave bands of high spectrum image is divided into multiple high correlation wave bands
Group, it is ensured that there is the 3rd set s in each high correlation band group3In at least one element；
For each high correlation band group, using the maximum element of information entropy in the group as the group reference wave band,
It regard other wave bands in the group as non-reference wave band.
Alternatively, the wave band that low correlation wave band is removed in all wave bands of high spectrum image is divided into multiple high correlations
The step of property band group, including：
Wave band to be grouped is used as using low correlation wave band is removed in all wave bands of high spectrum image；
Wave band to be grouped is divided into three high correlation band groups, wherein, the number of the first high correlation band group medium wave band
Amount is less than the 3rd preset parameter value T, and the quantity of the second high correlation band group medium wave band is equal to the 3rd preset parameter value T, the 3rd
The quantity of high correlation band group medium wave band, which is more than in the 3rd preset parameter value T, and each high correlation band group, has the 3rd
Set s3In at least one element.
Alternatively, step 02 includes：
N blocks will be divided into reference to wave band, the m blocks in n blocks are regard as area-of-interest；
N, m are natural number, and m is less than n.
Alternatively, the first coded system is the coded system that distributed compression perceives JSM-3 sparse models, i.e., all signals
Include public non-sparse part and distinctive sparse part；
Second coded system is the coded system that distributed compression perceives JSM-2 sparse models, i.e., all signals are included
Public sparse part and distinctive sparse part；
3rd coded system is the coded system that distributed compression perceives JSM-1 sparse models, i.e., all signals have identical
Sparsity structure.
Alternatively, methods described also includes：
Step 06：Receive the code stream after the compressed encoding that coding side is sent；
Step 07：Each group of area-of-interest is reconstructed using the first decoding process and multiple associated with area-of-interest
Residual image；
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process；
The image after the sparse transformation of low correlation wave band is reconstructed using the 3rd decoding process；
Step 08：According to the area-of-interest and multiple residual images associated with area-of-interest for reconstructing all groups,
Background area and multiple residual images associated with background area, the image of low correlation wave band, and then reconstruct high-spectrum
Picture.
Alternatively, step 08 includes：
For each group, it is added, obtains by area-of-interest, with the residual image that each is associated with the area-of-interest
The first area of each non-reference wave band；
It is added by background area, with the residual image that each is associated with the background area, obtains each non-reference ripple
The second area of section；
Area-of-interest is added with background area, the reference wave band of reconstruct is obtained；
The first area of each non-reference wave band is added with the second area of corresponding non-reference wave band, reconstructed
Non-reference wave band；
Reference wave band, the non-reference wave band and low correlation wave band of reconstruct according to reconstruct, reconstruct high spectrum image.
Second aspect, the present invention provides a kind of coding/decoding method based on method for compressing high spectrum image, including：
Receive the code stream of compressed encoding；
Each group of area-of-interest and multiple residual errors associated with area-of-interest are reconstructed using the first decoding process
Image；
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process；
The image after the sparse transformation of low correlation wave band is reconstructed using the 3rd decoding process；
According to the area-of-interest and multiple residual images associated with area-of-interest, background area for reconstructing all groups
And multiple residual images associated with background area, the image of low correlation wave band, reconstruct high spectrum image.
Alternatively, according to all groups of area-of-interest and multiple residual images associated with area-of-interest, background area
Domain and multiple residual images associated with background area, the image of low correlation wave band, the step of reconstructing high spectrum image, bag
Include：
For each group, it is added, obtains by area-of-interest, with the residual image that each is associated with the area-of-interest
The first area of each non-reference wave band；
It is added by background area, with the residual image that each is associated with the background area, obtains each non-reference ripple
The second area of section；
Area-of-interest and background area are combined, the reference wave band of reconstruct is obtained；
The first area of each non-reference wave band is added with the second area of corresponding non-reference wave band, reconstructed
Non-reference wave band；
Reference wave band, the non-reference wave band and low correlation wave band of reconstruct according to reconstruct, reconstruct high spectrum image.
Alternatively, each group of area-of-interest is reconstructed using the first decoding process and multiple associated with area-of-interest
Residual image, including：
Each group of area-of-interest is recovered using Hafman decoding mode, the joint weight perceived according to distributed compression
Structure method reconstructs multiple residual images associated with area-of-interest；
And/or,
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process,
Including：
The combined reconstruction method perceived according to distributed compression reconstructs background area and multiple associated with background area
Residual image；
And/or,
The image of low correlation wave band is reconstructed using the 3rd decoding process, including：
The combined reconstruction method perceived according to distributed compression reconstructs the low correlation image after sparse transformation.
The present invention has following beneficial effects based on the method for compressing high spectrum image that distributed compression is perceived：
First, the above method can carry out selection simultaneously to the area-of-interest and wave band interested of high spectrum image and carry
Take, it is ensured that the reservation of effective information in spectrum compression process.That is, the above method is based on spectrum correlation and entropy is combined
Waveband selection and packet are carried out, high entropy is selected and the high wave band of correlation is as wave band is referred to, to reference to the interested of wave band
Region, which carries out Lossless Compression, can preferably protect the important information of high spectrum image.
Second, the above method can filter out the wave band of high entropy low correlation, can prevent these wave bands from turning into reference
Wave band, improves the precision of reconstruct high spectrum image.
3rd, above method sampling process is simple, and observation process is random, during transmission code stream, even if producing few
The mistake of amount, nor affects on restructuring procedure, and cause error-resilient performance enhancing.
4th, the above method is classified to all wave bands of EO-1 hyperion, by the different disposal to different-waveband, can be with
While useful information is protected, compression efficiency is improved.
Brief description of the drawings
The process schematic that all wave bands of high spectrum image are grouped that Fig. 1 provides for one embodiment of the invention；
The schematic diagram that ROI region is selected in the reference wave band that Fig. 2 provides for one embodiment of the invention；
The schematic diagram for the method for compressing high spectrum image that Fig. 3 provides for one embodiment of the invention；
Fig. 4 is the coding based on spectrum correlation and the united distributed compression perception algorithm of entropy of one embodiment of the invention
Decoding process schematic diagram；
Fig. 5 a are the schematic diagram of the 45th wave band of original Terrain images；
Fig. 5 b are the schematic diagram of the 89th wave band of original Terrain images；
Fig. 5 c are the schematic diagram of the 45th wave band recovered using existing 3D-SPIHT modes；
Fig. 5 d are the schematic diagram of the 89th wave band recovered using existing 3D-SPIHT modes；
Fig. 5 e are the schematic diagram of the 45th wave band recovered using existing IOI-DCS modes；
Fig. 5 f are the schematic diagram of the 89th wave band recovered using existing IOI-DCS modes；
Fig. 5 g are the schematic diagram of the 45th wave band recovered using the CE-DCS modes of the present invention；
Fig. 5 h are the schematic diagram of the 89th wave band recovered using the CE-DCS modes of the present invention；
Fig. 6 a are the schematic diagram of the 20th wave band of original Cuprite images；
Fig. 6 b are the schematic diagram of the 90th wave band of original Cuprite images；
Fig. 6 c are the schematic diagram of the 20th wave band recovered using existing 3D-SPIHT modes；
Fig. 6 d are the schematic diagram of the 90th wave band recovered using existing 3D-SPIHT modes；
Fig. 6 e are the schematic diagram of the 20th wave band recovered using existing IOI-DCS modes；
Fig. 6 f are the schematic diagram of the 90th wave band recovered using existing IOI-DCS modes；
Fig. 6 g are the schematic diagram of the 20th wave band recovered using the CE-DCS modes of the present invention；
Fig. 6 h are the schematic diagram of the 90th wave band recovered using the CE-DCS modes of the present invention.
Embodiment
In order to preferably explain the present invention, in order to understand, below in conjunction with the accompanying drawings, by embodiment, to this hair
It is bright to be described in detail.
The compression method of high spectrum image is that all wave bands and all area of space of image are adopted mostly in the prior art
Take identical compression to handle, so unavoidably important information can be caused to lose.Even if the compression algorithm for having only a few considers height
The problem of spectrum picture information of interest retains, also just for one in wave band interested and area-of-interest, do not examine
Consider and retained both simultaneously.The method for compressing high spectrum image of the present invention is to area-of-interest in image and wave band interested
Selective extraction is all carried out, no longer for one of those, but to both retaining.For example, selecting high entropy first
Wave band, next selects the high wave band of correlation, using the two common wave band as referring to wave band, to reference to the interested of wave band
Region, which carries out Lossless Compression, can preferably protect the important information of high spectrum image, and error-resilient performance is strong.
Embodiment one
As shown in figure 1, the present embodiment provides a kind of method for compressing high spectrum image perceived based on distributed compression, the party
Method includes：
Step 01：For pending high spectrum image, based on each wave band correlation in high spectrum image and each wave band
All wave bands of high spectrum image are divided into a low correlation band group and multiple high correlation band groups by information entropy,
Each high correlation band group includes：One refers to wave band and multiple non-reference wave bands；
For example, the step 01 may include：
011st, the information entropy of all wave bands in the high spectrum image is obtained；It will be greater than the comentropy of the first default entropy
It is worth corresponding wave band composition first set s1；
012nd, the correlation coefficient r in the high spectrum image between all adjacent bands is obtained；It is determined that being preset more than first
Those wave bands are constituted second set s by corresponding two wave bands of each coefficient correlation of parameter value2；
013rd, by first set s1With second set s2Common factor be used as the 3rd set s3；By first set s1It is middle to remove the
Three set s3Element be used as a low correlation band group；
014th, the wave band that low correlation wave band is removed in all wave bands of high spectrum image is divided into multiple high correlations
Band group, so as to there is the 3rd set s in each high correlation band group3In at least one element；
015th, for each high correlation band group, the 3rd set s will be belonged in the group3Element in information entropy it is maximum
Element as the reference wave band of the group, regard other wave bands in the group as non-reference wave band.
Step 02：For each high correlation band group, according to default area-of-interest selection strategy, from referring to wave band
Middle determination area-of-interest, will be used as background area with reference to the wave band in wave band in addition to area-of-interest.
For example, n blocks will be divided into reference to wave band, the m blocks in n blocks are regard as area-of-interest；As shown in Fig. 2 will refer to
Wave band is divided into 9 pieces, regard 5 pieces of ground in 9 pieces as area-of-interest.
N, m are natural number, and m is less than n.
Step 03：For each non-reference wave band of each high correlation band group, referred to the group in wave band and feel emerging
Interesting region, background area carry out difference processing respectively, obtain each self-corresponding residual image.
Step 04：Using area-of-interest in each high correlation band group of the first coded system successively compressed encoding and many
The individual residual image associated with area-of-interest；
Using background area and multiple and background in each high correlation band group of the second coded system successively compressed encoding
The residual image of region association；
Image after the sparse transformation of low correlation band group is encoded using the 3rd coded system successively independent compression.
For example, the first coded system in the step can perceive the coding of JSM-3 sparse models for distributed compression
Mode, i.e., all signals include public non-sparse part and distinctive sparse part；
Second coded system can perceive the coded system of JSM-2 sparse models, i.e., all signal bags for distributed compression
Containing public sparse part and distinctive sparse part；
3rd coded system is the coded system that distributed compression perceives JSM-1 sparse models, i.e., all signals have identical
Sparsity structure.
Perception random observation measurement is compressed in the present embodiment to background area and residual image, background area is distributed
Higher sample rate, to residual image distribution compared with low sampling rate, keeps average sample rate essentially identical.
Step 05：The code stream of all compressed encodings is sent.
In the present embodiment, selective extraction can be carried out simultaneously to the area-of-interest and wave band interested of high spectrum image,
Ensure the reservation of the effective information in spectrum compression process.That is, the above method is based on spectrum correlation and entropy joint carries out wave band
Selection and packet, select high entropy and the high wave band of correlation is as wave band is referred to, and the area-of-interest with reference to wave band is carried out
Lossless Compression can preferably protect the important information of high spectrum image.
The wave band of high entropy low correlation can be filtered out in the present embodiment, can prevent these wave bands from turning into reference wave
Section, improves the precision of reconstruct high spectrum image.
In a kind of optional implementation, above-mentioned sub-step 014 may particularly include：
0141st, it regard the wave band that low correlation wave band is removed in all wave bands of high spectrum image as wave band to be grouped；
0142nd, wave band to be grouped is divided into three high correlation band groups, wherein, the first high correlation band group medium wave band
Quantity be less than the 3rd preset parameter value T, the quantity of the second high correlation band group medium wave band is equal to the 3rd preset parameter value T,
The quantity of 3rd high correlation band group medium wave band, which is more than in the 3rd preset parameter value T, and each high correlation band group, to be present
3rd set s3In at least one element.
In specific implementation process, the above method also includes following decoding process steps 06 described as follows to step 08：
Step 06：Receive the code stream after the compressed encoding that coding side is sent；
Step 07：Each group of area-of-interest is reconstructed using the first decoding process and multiple associated with area-of-interest
Residual image；
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process；
The image after the sparse transformation of low correlation wave band is reconstructed using the 3rd decoding process.
For example, each group of area-of-interest and multiple and region of interest is recovered using Hafman decoding mode
The residual image of domain association；
And/or, according to distributed compression sensing reconstructing mode reconstructed background region and it is multiple associated with background area it is residual
Difference image；The combined reconstruction method perceived according to distributed compression reconstructs the low correlation image after sparse transformation.
Step 08：According to all groups of area-of-interests and multiple residual images associated with area-of-interest, background area
Domain and multiple residual images associated with background area, the image of low correlation wave band, reconstruct high spectrum image.
In step 08, for each group, the residual plot associated by area-of-interest, with each with the area-of-interest
As being added, the first area of each non-reference wave band is obtained；
It is added by background area, with the residual image that each is associated with the background area, obtains each non-reference ripple
The second area of section；
Area-of-interest and background area are combined, the reference wave band of reconstruct is obtained；
The first area of each non-reference wave band is added with the second area of corresponding non-reference wave band, reconstructed
Non-reference wave band；
Reference wave band, the non-reference wave band and low correlation wave band of reconstruct according to reconstruct, reconstruct high spectrum image.
Above method sampling process is simple, and observation process is random, during transmission code stream, even if producing a small amount of mistake
By mistake, restructuring procedure is nor affected on, and causes error-resilient performance enhancing.
Embodiment two
Traditional Information Entropy is used for the wave band for selecting informative.These general wave bands all can by special treatment, for example,
In the technologies based on prediction such as such as Differential pulse code modulation (DPCM), selected wave band typically can all be taken as and be gone with reference to wave band
Predict other non-reference wave bands.And the precision predicted can be by the strong and weak shadow of correlation between selected wave band and all band of its in group
Ring, if the wave band selected according to Information Entropy and the remaining wave band correlation very little in group, then corresponding predicated error will
It is very big.Analyzed from the characteristic of high spectrum image, both correlation and entropy curve have similar between high spectrum image adjacent band
Trend, but correlation curve and entropy curve are not fully overlapped, and even opposite trend occur in the turning point of curve,
That is when some wave band has larger entropy, the correlations of correspondence and other wave bands might not very by force, and correlation it is strong when,
Entropy is also not necessarily larger.Higher entropy is possessed to these and but possesses wave band compared with low correlation, if continuing according to prior art
It is selected as clearly irrational with reference to wave band.
Therefore, the embodiment of the present invention is proposed based on spectrum correlation and the united waveband selection of entropy and using distributed pressure
The compression algorithm of compression method, i.e. CE-DCS algorithms, the information of interest of high spectrum image is protected with this.
The method for compressing high spectrum image perceived with distributed compression is protected based on information of interest, as shown in figure 3, including
Following steps：
Step A01：Original high spectrum image is grouped using each wave band spectrum correlation and comentropy, one is divided into
Low correlation band group and multiple high correlation band groups, the reference wave band of the group is selected from each high correlation band group
xr, non-reference wave band xn, as shown in Figure 1.
Step A01 includes following sub-steps：
Sub-step 1.1 obtains the information entropy of all wave bands in the high spectrum image；It will be greater than the first default entropy
The corresponding wave band composition first set s of information entropy1；
Entropy calculating is carried out firstly, for all wave bands of high spectrum image, the entropy of each wave band is obtained using formula
Size：
Wherein, p (i) is the probability for the pixel that gray value is i occur in high spectrum image, and E is comentropy.High spectrum image
There are many wave bands, use EjTo represent the comentropy of jth wave band, p is usedj(i) there is gray value in the image to represent jth wave band
For pixel i probability.
Secondly, maximum informational entropy E in all wave bands is found in high spectrum imagemax, calculate the average information entropy of all wave bands
V.Threshold alpha=V* β (wherein 1 ＜ β ＜ E are setmax/ V), if Ei＞ α, then be put into set s by i wave bands1；That is set s1In wave band
All it is the wave band of high entropy.
Sub-step 1.2：Select the strong wave band of all correlations of whole high spectrum image and put it into set s2；
For example, carrying out correlation analysis to all wave bands of high spectrum image, the phase relation between all adjacent bands is calculated
Number rk(m), maximizing rmaxWith average value rV.Threshold value R=r is setV* δ (wherein 1 ＜ δ ＜ rmax/rV)：
R in formulak(m) be kth wave band vector xk(data of each wave band can regard a column vector, x askRepresent
The expression-form of k-th of wave band column vector) in m elements, S is the number of pixel,It is kth ripple
The average value of all pixels in section image, if rk＞ R, then be put into set s by k and k+1 wave bands2。
Sub-step 1.3：Seek s1And s2Common factor s3, as shown in Figure 1；
For example, for Terrain images set β=1.09, R=0.993, for Cuprite images set β=
1.04, R=0.995, wave band initial option situation is as shown in Table 1 and Table 2：
Table 1：Waveband selection result of the Terrain images based on CE algorithms
Table 2：Waveband selection result of the Cuprite images based on CE algorithms
Sub-step 1.4：Remove in set s1In without in set s3In wave band (remove the ripple of high entropy low correlation
Section) and other low correlation wave bands, correlation packet is carried out to remaining wave band, that is, obtains multiple high correlation band groups；
Sub-step 1.5：Judge whether there is set s in every group3In wave band, if any then selection maximum entropy wave band be used as ginseng
Examine wave band；Otherwise the group and preceding a small group are merged into one group, the reference wave band using previous group is refers to wave band.
For example, calculating the correlation between adjacent band after the wave band of all wave bands removing low correlations in high spectrum image
Coefficient rk, and threshold value is set(wherein), if rk＞ R1, then wave band k and k+1 wave band is put into together
In a subset n, otherwise k+1 wave bands are put into subset (n+1),
Calculate the number of each subset medium wave band.The wave band number included in tentative standard subset is T, then including wave band
Subset of the number less than T is used as a group；Otherwise using T adjacent wave bands as one group, remaining last group of wave band number is less than T
When, the group is merged with previous group.
Step A02：For each high correlation band group, area-of-interest (ROI) is selected in reference to wave band, its
Remaining is background area.
In high spectrum image, information of interest is made up of two parts, is on the one hand wave band interested (Band Of
Interest, BOI), on the other hand it is area-of-interest (Region Of Interest, ROI).BOI's is proposed to
Information most abundant wave band is selected on the basis of original high-spectral data with certain criterion.ROI is in graphical analysis
The region constituted by many pixels of special consideration.Realize that the compression protected based on information of interest first will be selected and extracted
Go out BOI and ROI.Above-mentioned s3Wave band in set is exactly the BOI that each original image is selected, and each wave band includes more letters
Breath amount.
In the present embodiment, each BOI chosen image can be divided into some pieces of regions, each piece of region is entered
Line number, that block of selection comprising area-of-interest carries out lossless or near lossless compression to it, it is ensured that the reconstruct matter in this block region
Amount, larger compression ratio is then used to background area (BG regions).
In the present embodiment, one BOI is divided into equal-sized 9 pieces of regions for the convenience of experiment, selected most middle
That block (i.e. the 5th piece) ROI the most, remaining is then BG regions, and specific piecemeal situation is as shown in Figure 2.
Step A03：For each non-reference wave band of each high correlation band group, referred to the group in wave band and feel emerging
Interesting region, background area carry out difference processing respectively, obtain each self-corresponding residual image；
For example, according to step A01 packet, having in every group with reference to wave band xrWith non-reference wave band xn, by every group of ginseng
Examine wave band and non-reference wave band carries out difference, obtain residual imageI.e.
More sparse residual image is obtained by difference operationBecause high spectrum image in itself each wave band be it is not sparse,
But there is very strong correlation between wave band, so corresponding residual image is sparse, and then follow-up distributed compression can be carried out.
Step A04：According to the difference of significance level, different coded treatments are taken different piece；
For example, high spectrum image is only left reference picture (including ROI and BG regions) and residual plot after difference processing
Picture.ROI region is non-sparse image in the present embodiment, because high spectrum image has very strong spectrum correlation, so passing through difference
Point operation, residual image be sparse image, this allow for reference to wave band ROI region image and non-reference wave band it is corresponding should
Area image meets JSM-3 sparse models, i.e., be made up of non-sparse common portion and peculiar sparse part.
And refer to the BG regions in wave band by be after sparse transformation it is sparse, can as public sparse part, with reference to
Corresponding residual image meets JSM-2 sparse models, i.e., be made up of public sparse part and peculiar sparse part.
For in s1In without in s2In wave band and other low correlation wave bands, each of which passes through rarefaction representation
All it is that sparse, all signal all has a common sparse base afterwards, but the sparse coefficient of different signals is different, symbol
Close JSM-1 sparse models.The present embodiment carries out independent compression to each of which and handled.For different-waveband or region not
It is as shown in Figure 3 with processing mode.The coding and decoding process of each JSM model is as shown in Figure 4.
Sub-step 4.1：Perception random observation measurement is compressed to background area and residual image, background area is distributed
Compared with high sampling rate, to residual image distribution compared with low sampling rate, keep average sample rate constant；
For BG regions, be after sparse transformation with reference to the BG regions in wave band it is sparse, can be as public sparse
Part, because wave band has very strong correlation in group, so residual image can as the distinctive sparse part of each wave band,
This has just met JSM-2 sparse models：
Wherein：xjFor distributed primary signal, ψ is exactly public sparse base,And σjRespectively common portion and sparse part
Sparse coefficient, for the public sparse part of wave band, i.e., the reference wave band BG in every group is adopted using higher sample rate
Sample, so while reduced overall effect is ensured, can use lower sample rate residual image.
In compressed encoding, present embodiment will be applied to image with one kind and be easy to hard-wired measurement square
Battle array --- part Hadamard calculation matrix, hadamard matrix is made up of+1 and -1 element and meets AA '=nE (A ' is here
A transposition, E is unit square formation, and n is order of matrix number), the construction of hadamard matrix only needs to plus and minus calculation, and amount of calculation is small, surveys
Amount process need not take very big memory headroom.Its building method is：The hadamard matrix of N × N size is firstly generated,
Then it is random that M row vectors are chosen from the hadamard matrix, constitute the calculation matrix that a size is M × N.Due to Hadamard
Matrix is orthogonal matrix, therefrom takes the part hadamard matrix of the M × N sizes obtained after M rows and still has stronger non-phase
Close row and partial orthogonality, so compared with other certainty matrixes, measurement number required for the calculation matrix Exact Reconstruction compared with
It is few, that is to say, that under same measurement number, the reconstruction effect of part hadamard matrix is relatively good.
Each residual image independence by a M × N (M ＜ N) part Hadamard calculation matrix Φ, by it from a N-dimensional
Spatial sampling is to M dimension spaces, to the i-th wave band residual image, wherein 0 ＜ i ＜ k, sampling observation process is as follows：
The observation of obtained residual image is：
Openness very strong due to residual image, it is possible to less observation, i.e., lower sample rate is reconstructed
Residual image.
By carrying out identical processing to every group of wave band, by sparse transformation matrix, compressed sensing observing matrix, sight in sampling
Measured value is transferred to decoding end and rebuild.
Sub-step 4.2：Huffman lossless coding is taken to the ROI region in reference picture；
High spectrum image is non-sparse with reference to ROI region in wave band and the corresponding region of non-reference wave band, due to group
Interior wave band has very strong correlation, so the residual image of the corresponding non-reference wave band of ROI region is sparse, meets distribution
JSM-3 sparse models in formula compressed sensing, i.e., be made up of non-sparse common portion and peculiar sparse part.
xj=xc+sj=xc+ψυj, j ∈ { 1,2 ..., S },
Wherein xjRepresent the signal that j-th of band reception is arrived, υjFor the sparse coefficient of unique portion, they are different, xc
For not sparse common portion.For residual image, we use the observation procedure in sub-step 4.1 to measure, and are observed
Value.The method that Lossless Compression is used for ROI region, finds, the effect of Huffman encoding is more preferable by contrast.
Sub-step 4.3：Perception random observation measurement is compressed to low correlation wave band；
For being removed in sub-step 1.4 in set s1In without in set s3In wave band, i.e. high entropy low phase closes
Property wave band and other low correlation wave bands, they possess common sparsity structure, meet JSM-1 sparse models：
xj=ψ θj,j∈{1,2,...,S}
Wherein xjBand signal is represented, ψ is exactly public sparse base, sparse base uses wavelet transform (DWT), and θjFor
The different sparse coefficients of unlike signal.I.e. all signals all have a common sparse base, but different signal is dilute
Sparse coefficient is different, and carrying out independent compressed sensing random observation to each of which wave band measures.Observation procedure uses sub-step
Observation procedure in 4.1 is measured, and obtains observation.
Sub-step 4.4：To decoding end transmission Huffman encoding code stream, sparse transformation matrix, compressed sensing observing matrix, sight
The relevant information of the codings such as measured value；
Step A05：Different coding process is decoded accordingly in decoding end, combined reconstruction goes out high spectrum image；
In distributed compression perception theory, combined coding mode is used in coding for distributed signal, during reconstruct
Take combined reconstruction method.By taking the reconstruct of two sparse models of JSM-2 and JSM-3 as an example, encoded to high spectrum image
Shi Huixian is selected with reference to wave band and non-reference wave band, is then calculated both differences and is obtained residual image, according to referring to wave band
ROI region and the different sparse situation in BG regions, distributed compression is carried out using different sparse models.By the code after coding
It is streamed to after decoding end, first reconstructs correspondence reference picture and residual image, be finally added residual image with reference picture
Obtain non-reference picture.JSM-1 sparse models are that independent CS codings are carried out to each wave band in coding, decode combined reconstruction
Go out all low correlation wave bands.
Sub-step 5.1：Recovered using Hafman decoding with reference to corresponding area-of-interest in wave band；
Receive the Huffman encoding code stream of coding side transmission in decoding end, Huffman encoding code stream is reconstructed.Recover
Go out the area-of-interest with reference to wave band.
Sub-step 5.2：Utilize corresponding background area in distributed compression sensing reconstructing method reconstructed reference wave band and right
Answer the residual image in region；
The reconstructing method selection of the present embodiment is that base follows the trail of (Basis Pursuit, BP) method, according to step 4.1
The measured value obtained after sampled measurements is
Based on l1The compressed sensing signal reconstruction of norm passes through min | | x | |1,Carry out
Solve.
Sub-step 5.3：The wave band of low correlation is reconstructed using distributed compression sensing reconstructing method；
(Basis Pursuit, BP) method is equally followed the trail of in reconstruct for the wave band of low correlation with base, by changing into
l1The optimization problem of norm reconstructs original band image.
Sub-step 5.4：The reference band image of reconstruct is added with residual image and obtains non-reference band image；
Sub-step 5.5：By the reference band image and non-reference band image and all low correlation wave band phases of reconstruct
With reference to obtaining whole high-spectral data.
The above method removes that entropy is larger and the less wave band of correlation first, to prevent these wave bands to be selected as reference wave
Section, for these wave bands and other low correlation wave bands, meets JSM-1 models after rarefaction representation；Then will be remaining
Band grouping, and the wave band for making correlation stronger puts together, it is ensured that the wave band in each packet has stronger correlation
Property；Finally one is selected in each group and suitably refer to wave band, be that the difference of next step is prepared.It can thus be seen that base
Precision of prediction is improved in spectrum correlation and the united Hyperspectral image compression algorithm of comentropy (i.e. CE-DCS algorithms), is also increased
Processing to high entropy low correlation wave band, ensures larger compression ratio while important information is retained, therefore propose
CE-DCS algorithms have apparent advantage.
Embodiment three
Illustrate the superiority of present embodiment with reference to experimental data and experimental result：
It is object that Terrain and the panel height spectrum pictures of Cuprite two are chosen in experiment, and sparse base uses DWT, random observation square
Battle array is part hadamard matrix, and CS reconstructing methods are that base follows the trail of (Basis Pursuit, BP) method.
Waveband selection result is constant, and packet situation is as shown in Table 1 and Table 2.Sample rate SR=M/N × 100%, M is observation
The line number of matrix, N is the length of sparse rear primary signal.(Intel monokaryon 2.66GHz/32 bit manipulation systems under the same conditions
Internal memory 2GB), by 3D-SPIHT algorithms, IOI-DCS algorithms (algorithm that packet is fixed using wave band) and CE- in the present embodiment
DCS algorithms are contrasted.
Table 3 and table 4 sets forth this several method of the above being averaged in Terrain and Cuprite reconstruction image
Peak value signal to noise ratio (Average PSNR, APSNR) (average compression ratio scope 0.1bpp~0.5bpp).With IOI-DCS algorithm phases
Compare, under the same conditions, the overall Quality of recovery of image increases CE-DCS algorithms.BG regions in Terrain images
Average PSNR improves 0.74dB, and the average PSNR of non-reference wave band improves 0.47dB.BG regions is flat in Cuprite images
Equal PSNR improves 0.7dB, and the average PSNR of non-reference wave band improves 0.3dB, and this illustrates the embodiment of the present invention from objective
The feasibility and validity of the CE-DCS algorithms of middle proposition.
Table 3：Terrain recovers the average peak signal to noise ratio contrast of three kinds of methods of image
The Cuprite of table 4 recovers the average peak signal to noise ratio contrast of three kinds of methods of image
In addition, Fig. 5 and Fig. 6 does not give two images when sample rate is 0.2bpp, the subjective effect pair of reconstruction image
Than (Terrain chooses the 45th wave band and 89 wave bands, and Cuprite chooses the 20th wave band and the 90th wave band).Fig. 5 and Fig. 6 (c),
(d) it can be seen that the recovery image Quality of recovery based on 3D-SPIHT methods is worst in, it appears that fuzzy, many details are also failed to understand
Aobvious, in addition blocking effect is also obvious, and there is obvious striped the junction of block and block, have impact on the quality for recovering image.
And the recovery image based on IOI algorithms is relatively clear but is not fine, such as shown in Fig. 5 and Fig. 6 (e), (f).
As shown in Fig. 5 and Fig. 6 (g), (h), the recovery image that the algorithm based on CE is obtained is most clear under the same conditions
Clear, also without blocking effect, closest with original image, recovery effects are best in three kinds of methods.Reason is the choosing of CE wave bands
The power that grouping algorithm have also contemplated that the correlation between wave band while wave band information contained entropy size is considered is selected, is only had
Larger comentropy and the wave band for having larger correlation with other wave bands, which could turn into, refers to wave band, and the grouping process of CE algorithms is
The strong wave band of correlation is assigned to one group, this guarantees differential process error is smaller, so as to obtain good Quality of recovery.
To sum up experiment as can be seen that CE-DCS algorithms compared with traditional algorithm high spectrum image correlation is utilized more thoroughly,
Improve precision of prediction, also add the processing to high entropy low correlation wave band, retain important information while ensure compared with
Big compression ratio, thus the CE-DCS algorithms proposed have apparent advantage.
Example IV
The embodiment of the present invention provides a kind of coding/decoding method based on method for compressing high spectrum image, and method includes：
Step B01, the code stream for receiving compressed encoding；
Step B02, the area-of-interest for reconstructing using the first decoding process each group and multiple closed with area-of-interest
The residual image of connection；
Step B03,
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process；
The image after the sparse transformation of low correlation wave band is reconstructed using the 3rd decoding process.
In this step, each group of area-of-interest is recovered using Hafman decoding mode, according to distributed compression
The combined reconstruction method of perception reconstructs multiple residual images associated with area-of-interest；
According to base method for tracing compressed sensing reconstruct mode reconstructed background region and it is multiple associated with background area it is residual
Difference image；The combined reconstruction method perceived according to distributed compression reconstructs the low correlation image after sparse transformation.
In the present embodiment, the second decoding process and the 3rd decoding process can be the same or different.
Step B04, the area-of-interest according to all groups and multiple residual images associated with area-of-interest, background area
Domain and multiple residual images associated with background area, the image of low correlation wave band, reconstruct high spectrum image.
For example, for each group, the residual image associated by area-of-interest, with each with the area-of-interest
It is added, obtains the first area of each non-reference wave band；
It is added by background area, with the residual image that each is associated with the background area, obtains each non-reference ripple
The second area of section；
Area-of-interest is added with background area, the reference wave band of reconstruct is obtained；
The first area of each non-reference wave band is added with the second area of corresponding non-reference wave band, reconstructed
Non-reference wave band；
Reference wave band, the non-reference wave band and low correlation wave band of reconstruct according to reconstruct, reconstruct high spectrum image.
Method in the present embodiment can improve the precision of reconstruct high spectrum image.Above-mentioned compression method sampling process is simple,
And observation process is random, during transmission code stream, even if producing a small amount of mistake, restructuring procedure is nor affected on, and cause anti-
Error performance strengthens.
Finally it should be noted that：Above-described embodiments are merely to illustrate the technical scheme, rather than to it
Limitation；Although the present invention is described in detail with reference to the foregoing embodiments, it will be understood by those within the art that：
It can still modify to the technical scheme described in previous embodiment, or which part or all technical characteristic are entered
Row equivalent substitution；And these modifications or substitutions, the essence of appropriate technical solution is departed from various embodiments of the present invention technical side
The scope of case.
Claims (10)
1. a kind of method for compressing high spectrum image perceived based on distributed compression, it is characterised in that including：
Step 01：For pending high spectrum image, the strong and weak and each wave band of the correlation based on each wave band in high spectrum image
Information entropy, all wave bands of high spectrum image are divided into a low correlation band group and multiple high correlation wave bands
Group, each high correlation band group includes：The reference wave band and multiple non-reference wave bands of one high entropy；
Step 02：For each high correlation band group, according to default area-of-interest selection strategy, from reference to true in wave band
Determine area-of-interest, background area will be used as with reference to the region in wave band in addition to area-of-interest；
Step 03：For each non-reference wave band of high correlation band group, area-of-interest in wave band, the back of the body are referred to the group
Scene area carries out difference processing respectively, and acquisition corresponds respectively to the area-of-interest and the residual image of background area；
Step 04：Using area-of-interest in each high correlation band group of the first coded system successively compressed encoding and it is multiple with
The residual image of area-of-interest association；
Using background area and multiple and background area in each high correlation band group of the second coded system successively compressed encoding
The residual image of association；
Image after the sparse transformation of low correlation band group is encoded using the 3rd coded system successively independent compression；
Step 05：The code stream of all compressed encodings is sent.
2. according to the method described in claim 1, it is characterised in that step 01 includes：
Obtain the information entropy of all wave bands in the high spectrum image；It will be greater than the corresponding wave band composition of the first default entropy the
One set s1；
Obtain the correlation coefficient r between all adjacent bands in the high spectrum image；It is determined that more than the first preset parameter value
Those wave bands are constituted second set s by corresponding two wave bands of each coefficient correlation2；
By first set s1With second set s2Common factor be used as the 3rd set s3；By first set s1The 3rd set s of middle removal3's
Element is used as one high entropy low correlation band group；
The wave band that low correlation wave band is removed in all wave bands of high spectrum image is divided into multiple high correlation band groups, protected
Demonstrate,prove and there is the 3rd set s in each high correlation band group3In at least one element；
For each high correlation band group, using the maximum element of information entropy in the group as the group reference wave band, by this
Other wave bands are used as non-reference wave band in group.
3. method according to claim 2, it is characterised in that low correlation will be removed in all wave bands of high spectrum image
The step of wave band of wave band is divided into multiple high correlation band groups, including：
Wave band to be grouped is used as using low correlation wave band is removed in all wave bands of high spectrum image；
Wave band to be grouped is divided into three high correlation band groups, wherein, the quantity of the first high correlation band group medium wave band is small
In the 3rd preset parameter value T, the quantity of the second high correlation band group medium wave band is equal to the 3rd preset parameter value T, the 3rd high phase
The quantity of closing property band group medium wave band, which is more than in the 3rd preset parameter value T, and each high correlation band group, has the 3rd set
s3In at least one element.
4. according to the method described in claim 1, it is characterised in that step 02 includes：
N blocks will be divided into reference to wave band, the m blocks in n blocks are regard as area-of-interest；
N, m are natural number, and m is less than n.
5. according to any described method of Claims 1-4, it is characterised in that the first coded system perceives for distributed compression
The coded system of JSM-3 sparse models, i.e., all signals include public non-sparse part and distinctive sparse part；
Second coded system is the coded system that distributed compression perceives JSM-2 sparse models, i.e., all signals are comprising public
Sparse part and distinctive sparse part；
3rd coded system is the coded system that distributed compression perceives JSM-1 sparse models, i.e., all signals have identical dilute
Dredge structure.
6. according to any described method of Claims 1-4, it is characterised in that methods described also includes：
Step 06：Receive the code stream after the compressed encoding that coding side is sent；
Step 07：Using the first decoding process reconstruct each group area-of-interest and it is multiple associated with area-of-interest it is residual
Difference image；
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process；
The image after the sparse transformation of low correlation wave band is reconstructed using the 3rd decoding process；
Step 08：According to the area-of-interest and multiple residual images associated with area-of-interest, background for reconstructing all groups
Region and multiple residual images associated with background area, the image of low correlation wave band, and then reconstruct high spectrum image.
7. method according to claim 6, it is characterised in that step 08 includes：
For each group, it is added, obtains each by area-of-interest, with the residual image that each is associated with the area-of-interest
The first area of non-reference wave band；
It is added by background area, with the residual image that each is associated with the background area, obtains each non-reference wave band
Second area；
Area-of-interest is added with background area, the reference wave band of reconstruct is obtained；
The first area of each non-reference wave band is added with the second area of corresponding non-reference wave band, the non-of reconstruct is obtained
With reference to wave band；
Reference wave band, the non-reference wave band and low correlation wave band of reconstruct according to reconstruct, reconstruct high spectrum image.
8. a kind of coding/decoding method based on method for compressing high spectrum image, it is characterised in that including：
Receive the code stream of compressed encoding；
Each group of area-of-interest and multiple residual images associated with area-of-interest is reconstructed using the first decoding process；
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process；
The image after the sparse transformation of low correlation wave band is reconstructed using the 3rd decoding process；
According to the area-of-interest and multiple residual images associated with area-of-interest that reconstruct all groups, background area and many
The individual residual image associated with background area, the image of low correlation wave band, reconstruct high spectrum image.
9. method according to claim 8, it is characterised in that according to all groups of area-of-interests and it is multiple with it is interested
Residual image, background area and multiple residual images associated with background area, the image of low correlation wave band that region is associated,
The step of reconstructing high spectrum image, including：
For each group, it is added, obtains each by area-of-interest, with the residual image that each is associated with the area-of-interest
The first area of non-reference wave band；
It is added by background area, with the residual image that each is associated with the background area, obtains each non-reference wave band
Second area；
Area-of-interest is added with background area, the reference wave band of reconstruct is obtained；
The first area of each non-reference wave band is added with the second area of corresponding non-reference wave band, the non-of reconstruct is obtained
With reference to wave band；
Reference wave band, the non-reference wave band and low correlation wave band of reconstruct according to reconstruct, reconstruct high spectrum image.
10. method according to claim 8 or claim 9, it is characterised in that
Each group of area-of-interest and multiple residual images associated with area-of-interest is reconstructed using the first decoding process,
Including：
Each group of area-of-interest is recovered using Hafman decoding mode, the combined reconstruction side perceived according to distributed compression
Method reconstructs multiple residual images associated with area-of-interest；
And/or,
Each group of background area and multiple residual images associated with background area are reconstructed using the second decoding process, wrapped
Include：
The combined reconstruction method perceived according to distributed compression reconstructs background area and multiple residual errors associated with background area
Image；
And/or,
The image of low correlation wave band is reconstructed using the 3rd decoding process, including：
The combined reconstruction method perceived according to distributed compression reconstructs the low correlation image after sparse transformation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710280701.9A CN107124612B (en) | 2017-04-26 | 2017-04-26 | Method for compressing high spectrum image based on distributed compression perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710280701.9A CN107124612B (en) | 2017-04-26 | 2017-04-26 | Method for compressing high spectrum image based on distributed compression perception |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107124612A true CN107124612A (en) | 2017-09-01 |
| CN107124612B CN107124612B (en) | 2019-06-14 |
Family
ID=59726358
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710280701.9A Expired - Fee Related CN107124612B (en) | 2017-04-26 | 2017-04-26 | Method for compressing high spectrum image based on distributed compression perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107124612B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108347612A (en) * | 2018-01-30 | 2018-07-31 | 东华大学 | A kind of monitored video compression and reconstructing method of view-based access control model attention mechanism |
| CN112241982A (en) * | 2019-07-18 | 2021-01-19 | 杭州海康威视数字技术股份有限公司 | Image processing method and device and machine-readable storage medium |
| CN113222835A (en) * | 2021-04-22 | 2021-08-06 | 海南大学 | Remote sensing full-color and multi-spectral image distributed fusion method based on residual error network |
| WO2022166363A1 (en) * | 2021-02-07 | 2022-08-11 | 浙江师范大学 | Neighboring subspace division-based hyperspectral imaging band selection method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8429153B2 (en) * | 2010-06-25 | 2013-04-23 | The United States Of America As Represented By The Secretary Of The Army | Method and apparatus for classifying known specimens and media using spectral properties and identifying unknown specimens and media |
| CN103780904A (en) * | 2014-01-23 | 2014-05-07 | 西安电子科技大学 | Hyperspectral image self-adaption lossless predictive encoding system and method based on edge information |
| CN104408751A (en) * | 2014-10-27 | 2015-03-11 | 中国科学院自动化研究所 | Hyperspectral image in-orbit compression method |
| CN105825530A (en) * | 2016-03-21 | 2016-08-03 | 辽宁师范大学 | Coastal zone hyperspectral image distributed lossy coding and decoding method based on region of interest |
| CN106529508A (en) * | 2016-12-07 | 2017-03-22 | 西安电子科技大学 | Local and non-local multi-feature semantics-based hyperspectral image classification method |
-
2017
- 2017-04-26 CN CN201710280701.9A patent/CN107124612B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8429153B2 (en) * | 2010-06-25 | 2013-04-23 | The United States Of America As Represented By The Secretary Of The Army | Method and apparatus for classifying known specimens and media using spectral properties and identifying unknown specimens and media |
| CN103780904A (en) * | 2014-01-23 | 2014-05-07 | 西安电子科技大学 | Hyperspectral image self-adaption lossless predictive encoding system and method based on edge information |
| CN104408751A (en) * | 2014-10-27 | 2015-03-11 | 中国科学院自动化研究所 | Hyperspectral image in-orbit compression method |
| CN105825530A (en) * | 2016-03-21 | 2016-08-03 | 辽宁师范大学 | Coastal zone hyperspectral image distributed lossy coding and decoding method based on region of interest |
| CN106529508A (en) * | 2016-12-07 | 2017-03-22 | 西安电子科技大学 | Local and non-local multi-feature semantics-based hyperspectral image classification method |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108347612A (en) * | 2018-01-30 | 2018-07-31 | 东华大学 | A kind of monitored video compression and reconstructing method of view-based access control model attention mechanism |
| CN108347612B (en) * | 2018-01-30 | 2020-09-15 | 东华大学 | Monitoring video compression and reconstruction method based on visual attention mechanism |
| CN112241982A (en) * | 2019-07-18 | 2021-01-19 | 杭州海康威视数字技术股份有限公司 | Image processing method and device and machine-readable storage medium |
| WO2022166363A1 (en) * | 2021-02-07 | 2022-08-11 | 浙江师范大学 | Neighboring subspace division-based hyperspectral imaging band selection method and system |
| CN113222835A (en) * | 2021-04-22 | 2021-08-06 | 海南大学 | Remote sensing full-color and multi-spectral image distributed fusion method based on residual error network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107124612B (en) | 2019-06-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107124612B (en) | Method for compressing high spectrum image based on distributed compression perception | |
| CN113516601A (en) | Image restoration technology based on deep convolutional neural network and compressed sensing | |
| CN103546759A (en) | Image compression coding method based on combination of wavelet packets and vector quantization | |
| Yadav et al. | A review on image compression techniques | |
| Mander et al. | An improved image compression-decompression technique using block truncation and wavelets | |
| George et al. | Image compression based on wavelet, polynomial and Quadtree | |
| CN105825530B (en) | Littoral zone high spectrum image distribution lossy coding and coding/decoding method based on area-of-interest | |
| Sadkhan | A proposed image compression technique based on DWT and predictive techniques | |
| CN108573509B (en) | Self-adaptive compressed sensing image coding method based on measurement domain significance detection model | |
| Hagag et al. | HyperCast: hyperspectral satellite image broadcasting with band ordering optimization | |
| Kekre et al. | Image compression based on hybrid wavelet transform generated using orthogonal component transforms of different sizes | |
| Midha et al. | Analysis of RGB and YCbCr color spaces using wavelet transform | |
| CN104486631B (en) | A kind of remote sensing image compression method based on human eye vision Yu adaptive scanning | |
| Panda et al. | Competency assessment of image compression in the lossy and lossless domain | |
| CN106131575A (en) | The method for compressing image combined with Chinese remainder theorem based on wavelet transformation | |
| Hu et al. | Colour image coding scheme using absolute moment block truncation coding and block prediction technique | |
| Jagadeesh et al. | Linear adaptive global node-tree filters based SPIHT MR image codec | |
| Thayammal et al. | Multispectral band image compression using adaptive wavelet transform-tetrolet transform | |
| Lakshmi et al. | Gaussian Restoration pyramid: Application of image restoration to Laplacian pyramid compression | |
| CN114422787B (en) | Image pseudo-analog wireless transmission method based on residual layer blocking DCT transformation | |
| Rostam et al. | Medical Image Application by Hybrid Transform Coding Scheme | |
| Verma et al. | Comparative Study and Analysis of DWT-SPIHT with DWT-EZW Method for Image Compression | |
| Gopi et al. | REGION OF INTEREST BASED NOVEL CODING TECHNIQUE FOR EFFICIENT COMPRESSION AND TRANSMISSION | |
| Vargic et al. | Hierarchical mask creation for intelligent image coding using saliency maps | |
| Verma et al. | Performance Analysiis and Comparison between Different Image Compression Techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190614 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |