CN112399182B - Single-frame infrared image hybrid compression method and system - Google Patents
Single-frame infrared image hybrid compression method and system Download PDFInfo
- Publication number
- CN112399182B CN112399182B CN202011088307.3A CN202011088307A CN112399182B CN 112399182 B CN112399182 B CN 112399182B CN 202011088307 A CN202011088307 A CN 202011088307A CN 112399182 B CN112399182 B CN 112399182B
- Authority
- CN
- China
- Prior art keywords
- image
- target
- foreground
- frequency
- similarity
- 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.)
- Active
Links
- 238000007906 compression Methods 0.000 title claims abstract description 97
- 230000006835 compression Effects 0.000 title claims abstract description 95
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000001914 filtration Methods 0.000 claims abstract description 66
- 239000013598 vector Substances 0.000 claims description 62
- 238000004364 calculation method Methods 0.000 claims description 22
- 239000011159 matrix material Substances 0.000 claims description 21
- 238000011084 recovery Methods 0.000 claims description 21
- 238000009792 diffusion process Methods 0.000 claims description 12
- 238000009499 grossing Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 230000017105 transposition Effects 0.000 claims description 5
- 230000003287 optical effect Effects 0.000 abstract description 7
- 230000005540 biological transmission Effects 0.000 description 10
- 238000004458 analytical method Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 7
- 238000013144 data compression Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000006837 decompression Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
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
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention relates to a single-frame infrared image hybrid compression method and a single-frame infrared image hybrid compression system. For an infrared source image, firstly carrying out lossy compression to obtain a lossy compression code stream and a lossy background, carrying out difference between a source image and the background to obtain a high-frequency foreground HF, carrying out non-uniform noise filtering on the HF to generate an image N, and simultaneously initializing a template M; the same operation is carried out on the target image T to obtain a target high-frequency foreground, and a structure descriptor K of the target image T is generatedT. Scanning N by taking pixel as unit, intercepting a sub-graph N with the same size as T by taking the current pixel as an origin in each scanning processiCalculating NiStructural descriptor of (1) and KTWhen the similarity is less than the threshold value, the M is compared with the NiThe corresponding region is set to 1 in its entirety. And after the scanning is finished, performing AND operation on the HF and the M, performing lossless compression, and combining the lossy and lossless compression code streams. The invention can effectively reduce the data scale of the single-frame infrared image and simultaneously reserve the optical characteristics of the infrared target.
Description
Technical Field
The invention relates to the field of single-frame infrared image compression, in particular to a single-frame infrared image hybrid compression method and a single-frame infrared image hybrid compression system.
Background
For some infrared imaging devices installed on highly dynamic equipment, a large amount of infrared data is generated due to the fact that images can be taken in tens to hundreds of frames per second, and for various reasons, the large-scale infrared data must be transmitted to a data center for analysis processing in real time through a wireless channel with the transmission rate of only several megabits. Meanwhile, considering the requirement of infrared target characteristic analysis, the infrared image data is not allowed to lose high-frequency details in the processing process, and therefore, the adoption of an image lossless compression technology to reduce the data scale becomes a preferred scheme of infrared data transmission preprocessing in theory. However, limited to the shannon limit, the data compression rate of the full lossless data compression technology, such as lossless JPEG2000, lossless JPEG-LS, etc., is still low, and thus the requirement of real-time transmission of large-scale infrared data through a low-rate channel cannot be met. On the other hand, although the lossy compression technique can greatly reduce the data size, the high-frequency details of the image are lost, so that the requirement of infrared target characteristic analysis cannot be essentially met. Therefore, the advantages of lossy compression and lossless compression are combined to perform mixed compression on source data, and on the premise of preserving the optical characteristics of the infrared target to be analyzed, the data compression rate is effectively improved to meet the requirement of transmitting the infrared image in real time by a low-speed channel, so that the method becomes an important direction for researching the infrared video data compression algorithm at present.
The article of academic parlance of Hongning, the article of academic parlance of Liu's article of academic parlance ' application research of ROI in remote sensing image coding transmission ', the article of academic parlance ' research of visual measurement image lossless ROI coding technology ', the article of academic parlance ' code rate control method for satellite image compression ', the article of Ningshigang's et al, the article of Queen's et al, the article of Wang's et al, the study of medical image segmentation and compression method based on optimized ROI, the invention patent of Huo hong Wei et al, circular region of interest in digital image and its compression algorithm ', the invention patent of Guihua et al, an ROI compression method of ship target slice image, etc. respectively proposes a method for extracting region of interest (ROI) in source image, namely a method for positioning and extracting an image region containing a target to be analyzed, and then uses the support of standard ROI block division priority coding technology such as JPEG2000, and the mixed compression is realized for the source image to a certain extent. The invention patents of Guo Ministry et al, namely medical image ROI compression method based on lifting wavelet and PCNN, the invention patents of Jiayi et al, namely ROI-based video rapid compression method, high-definition video system and 4K video system, the invention patents of Moyiarmy, namely CT influence interested region compression and quality evaluation method, and the like also respectively mention the positioning and related compression methods of ROI.
The related compression method has the following technical limitations:
1. part of methods depend on the existing image compression standards, such as lossless JPEG2000, lossless JPEG-LS and the like, so that certain flexibility is lacked;
2. the lossless compression object of the method is complete data in a calibration ROI area, if the data in the ROI area is non-sparse or poor in color tone continuity, the lossless compression ratio is actually low under the restriction of the Shannon limit;
3. the method does not consider the characteristics that most of energy of the infrared image is concentrated in a low-frequency interval, non-uniform noise is high and the like, so that the requirements of high compression rate of infrared image data and recovery of image infrared target characteristic analysis cannot be met;
4. some methods, such as jiamai patent of invention, adopt the difference method of the interframes to improve the compression rate greatly, but to the bad wireless communication environment, may have the problem that the transmission error rate is higher, therefore there is unrecoverable error in a frame data transmission, may influence the image recovery of the subsequent several frames.
Disclosure of Invention
The invention aims to provide a single-frame infrared image hybrid compression method and a single-frame infrared image hybrid compression system, which can better meet the requirements of real-time transmission of infrared video data and optical characteristic analysis of an infrared target of a low-speed wireless channel on the basis of effectively improving the infrared data compression ratio.
In order to achieve the purpose, the invention provides the following scheme:
a single-frame infrared image hybrid compression method comprises the following steps:
acquiring an infrared source image and a target image, and initializing a foreground template with the same size as the infrared source image;
lossy compression is carried out on the target image to obtain a target recovery image;
subtracting the target restored image from the target image to obtain a target difference high-frequency foreground image, and calculating a structure descriptor of the target difference high-frequency foreground;
lossy compression is carried out on the infrared source image to obtain a background compressed code stream and a restored image of the background compressed code stream, and the restored image of the background compressed code stream is an infrared source restored image;
subtracting the infrared source recovery image from the infrared source image to obtain an infrared source difference high-frequency foreground image;
filtering the infrared source difference high-frequency foreground image to obtain a filtered image;
scanning the filtering image according to a set pixel, and gradually intercepting a sub-image which takes a current pixel as an origin point at the upper left corner and has the same window scale as the target image;
calculating a structure descriptor of the subgraph;
calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph;
judging whether the similarity is smaller than a similarity threshold value;
if the similarity is smaller than a similarity threshold value, setting the corresponding area of the foreground template and the sub-image as 1;
judging whether the filtering image is traversed or not;
if the similarity is larger than or equal to a similarity threshold, directly judging whether the filtering image is traversed or not;
if the filtering image is traversed, performing AND operation on the foreground template and the infrared source differential high-frequency foreground image to obtain a high-frequency foreground image containing a target image;
lossless compression is carried out on the high-frequency foreground image containing the target image, and a foreground compression code stream is obtained;
merging the foreground compressed code stream and the background compressed code stream to obtain a mixed compressed code stream;
and if the filtering image is not traversed, returning to scanning the filtering image according to the set pixels, and gradually intercepting sub-images which take the current pixel as an original point at the upper left corner and have the same window size as the target image.
Optionally, the structural descriptor of the target differential high-frequency foreground adopts a local steering kernel, and a calculation formula of the local steering kernel is as follows:
wherein p is2The number of pixels contained in the sub-picture or the target image; h is a global smoothing parameter; matrix CiIs a covariance matrix of the spatial gradient vectors; x is the number ofjIs the coordinate of the center pixel point of the sub-image or the target image; x is the number ofiThe coordinates of other pixel points in the sub-image or the target image.
Optionally, the filtering the infrared source differential high-frequency foreground image to obtain a filtered image specifically includes:
filtering the infrared source difference high-frequency foreground image by adopting an anisotropic diffusion filtering method to obtain a filtered image, wherein the anisotropic diffusion filtering method adopts an iterative equation for filtering, and the iterative equation is as follows:
wherein ItIs the image at time t, It+1Is an image at the time of t +1 after one iteration; λ is a smoothing coefficient;
the divergence of the pixel (x, y) point in the four directions of east, west, south and north at the time t is respectively expressed as follows:
Sx，y、Ex，y、Nx，y、Wx，yrespectively representing the thermal conductivity in four directions,the specific calculation formula is as follows:
wherein k is a heat conduction correlation coefficient, and the larger the value of k is, the less easy the edge is kept in the iteration process.
Optionally, the calculating the similarity between the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph specifically includes:
using a formula
Calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph;
where ρ isiThe cosine similarity of two vectors determined by the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph, for the structure descriptor matrix K with Y columns and each column containing X elements, the structure vector K containing X X Y elements is constructed in a way that the adjacent column vectors are connected end to end and in the sequence of the column vectors, for this reason, the cosine similarity of the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph is defined as follows:
wherein (·)' represents the transposition operation of the vector, | | · | | | represents the vector modulo operation, κTStructural descriptor K for passing differential high-frequency foregroundTStructural vector of construction, κNiFor structure descriptor K by subgraphNiThe constructed structure vector.
A single frame infrared image hybrid compression system, comprising:
the image acquisition module is used for acquiring an infrared source image and a target image and initializing a foreground template with the same size as the infrared source image;
the first lossy compression module is used for carrying out lossy compression on the target image to obtain a target recovery image;
a structure descriptor determining module of the target difference high-frequency foreground image, which is used for subtracting the target recovery image from the target image to obtain a target difference high-frequency foreground image and calculating a structure descriptor of the target difference high-frequency foreground;
the second lossy compression module is used for lossy compressing the infrared source image to obtain a background compressed code stream and a recovered image of the background compressed code stream, wherein the recovered image of the background compressed code stream is an infrared source recovered image;
the infrared source difference high-frequency foreground image determining module is used for subtracting the infrared source recovery image from the infrared source image to obtain an infrared source difference high-frequency foreground image;
the filtering image determining module is used for carrying out filtering processing on the infrared source difference high-frequency foreground image to obtain a filtering image;
a subgraph determining module, which is used for scanning the filtering image according to the set pixel and gradually intercepting the subgraph which takes the current pixel as the origin of the upper left corner and has the same window scale as the target image;
the structure descriptor determining module of the subgraph is used for calculating the structure descriptor of the subgraph;
the similarity calculation module is used for calculating the similarity between the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph;
the first judgment module is used for judging whether the similarity is smaller than a similarity threshold value;
the region setting module is used for setting the region corresponding to the foreground template and the sub-image as 1 when the similarity is smaller than the similarity threshold;
the second judging module is used for judging whether the filtering image is completely traversed after the area corresponding to the foreground template and the sub-image is set to be 1 or judging whether the filtering image is completely traversed when the similarity is larger than or equal to the similarity threshold;
the high-frequency foreground image determining module is used for performing AND operation on the foreground template and the infrared source differential high-frequency foreground image to obtain a high-frequency foreground image containing a target image when the filter image is traversed;
the lossless compression module is used for carrying out lossless compression on the high-frequency foreground image containing the target image to obtain a foreground compressed code stream;
a mixed compressed code stream determining module, configured to combine the foreground compressed code stream and the background compressed code stream to obtain a mixed compressed code stream;
and the return module is used for returning to the subgraph determination module when the filtering image is not traversed.
Optionally, the structural descriptor of the target differential high-frequency foreground adopts a local steering kernel, and a calculation formula of the local steering kernel is as follows:
wherein p is2The number of pixels contained in the sub-picture or the target image; h is a global smoothing parameter; matrix CiIs a covariance matrix of the spatial gradient vectors; x is the number ofjIs the coordinate of the center pixel point of the sub-image or the target image; x is the number ofiThe coordinates of other pixel points in the sub-image or the target image.
Optionally, the filtered image determining module specifically includes:
the filtered image determining unit is used for filtering the infrared source difference high-frequency foreground image by adopting an anisotropic diffusion filtering method to obtain a filtered image, wherein the anisotropic diffusion filtering method adopts an iterative equation for filtering, and the iterative equation is as follows:
wherein ItIs the image at time t, It+1Is subjected to an iterationAn image at a later time t + 1; λ is a smoothing coefficient;
the divergence of the pixel (x, y) point in the four directions of east, west, south and north at the time t is respectively expressed as follows:
Sx，y、Ex，y、Nx，y、Wx，ythe heat conductivity coefficients in four directions are respectively represented, and the specific calculation formula is as follows:
wherein k is a heat conduction correlation coefficient, and the larger the value of k is, the less easy the edge is kept in the iteration process.
Optionally, the similarity calculation module specifically includes:
a similarity calculation unit for employing a formula
Calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph;
where ρ isiThe cosine similarity of two vectors determined by the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph, for the structure descriptor matrix K with Y columns and each column containing X elements, the structure vector K containing X X Y elements is constructed in a way that the adjacent column vectors are connected end to end and in the sequence of the column vectors, for this reason, the cosine similarity of the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph is defined as follows:
wherein(-) represents the transpose operation of the vector, | | - | | represents the vector modulo operation, κTStructure vector, κ, constructed from a Structure descriptor KT differentiating the high frequency perspectivesNiFor structure descriptor K by subgraphNiThe constructed structure vector.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
1. according to the characteristic that most energy of the infrared image is basically distributed in a low-frequency area, a mode of combining background lossy compression and high-frequency foreground lossless compression is adopted, the compression ratio is effectively improved, and the requirement of transmitting the infrared video in real time through a low-speed wireless channel is met;
2. a lossless compression mode is adopted for the high-frequency foreground containing the optical characteristics of the infrared target, and the requirement for restoring the characteristic analysis of the image infrared target is met.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a single-frame infrared image hybrid compression method according to the present invention;
FIG. 2 is a schematic diagram of a hybrid codec decompression process;
fig. 3 is a structural diagram of a single-frame infrared image hybrid compression system according to the present invention.
Fig. 4 is a schematic flow chart of a single-frame infrared image hybrid compression method according to embodiment 1 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a single-frame infrared image hybrid compression method and a single-frame infrared image hybrid compression system, which can better meet the requirements of real-time transmission of infrared video data and optical characteristic analysis of an infrared target of a low-speed wireless channel on the basis of effectively improving the infrared data compression ratio.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Aiming at an application scene and a related research technology, the invention provides a single-frame infrared image mixed compression method based on the combination of lossy and lossless compression technologies on the basis of fully considering the typical characteristics of an infrared image, the transmission characteristics of a low-speed wireless channel and the requirement of infrared target optical characteristic analysis. The method can effectively improve the compression ratio of the infrared data, and simultaneously better meet the requirements of real-time transmission of infrared video data of a low-speed wireless channel and optical characteristic analysis of an infrared target.
As shown in fig. 1, a single-frame infrared image hybrid compression method includes:
step 101: acquiring an infrared source image and a target image, and initializing a foreground template with the same size as the infrared source image.
Step 102: and carrying out lossy compression on the target image to obtain a target recovery image.
Step 103: and subtracting the target restored image from the target image to obtain a target difference high-frequency foreground image, and calculating a structure descriptor of the target difference high-frequency foreground.
The structural descriptor of the target differential high-frequency foreground adopts a local steering kernel, and the calculation formula of the local steering kernel is as follows:
wherein，p2The number of pixels contained in the sub-picture or the target image; h is a global smoothing parameter; matrix CiIs a covariance matrix of the spatial gradient vectors; x is the number ofjIs the coordinate of the center pixel point of the sub-image or the target image; x is the number ofiThe coordinates of other pixel points in the sub-image or the target image. Center pixel point xjThe LSK of (a) is a matrix of dimension P × P, and the specific calculation process is as follows:
1) for an image with the size of P multiplied by P, solving gradient vectors of all pixels in the image;
2) calculating its covariance matrix Ci。
3) LSK is calculated according to the above formula.
Step 104: and lossy compression is carried out on the infrared source image to obtain a background compressed code stream and a recovered image of the background compressed code stream, wherein the recovered image of the background compressed code stream is an infrared source recovered image.
Step 105: and subtracting the infrared source recovery image from the infrared source image to obtain an infrared source difference high-frequency foreground image.
Step 106: filtering the infrared source difference high-frequency foreground image to obtain a filtered image, and specifically comprising the following steps:
filtering the infrared source difference high-frequency foreground image by adopting an anisotropic diffusion filtering method to obtain a filtered image, wherein the anisotropic diffusion filtering method adopts an iterative equation for filtering, and the iterative equation is as follows:
wherein ItIs the image at time t, It+1Is an image at the time of t +1 after one iteration; λ is a smoothing coefficient;
the divergence of the pixel (x, y) point in the four directions of east, west, south and north at the time t is respectively expressed as follows:
Sx，y、Ex，y、Nx，y、Wx，ythe heat conductivity coefficients in four directions are respectively represented, and the specific calculation formula is as follows:
wherein k is a heat conduction correlation coefficient, and the larger the value of k is, the less easy the edge is kept in the iteration process. The reasonable values of k and λ can be obtained through experiments, and if k is 18 and λ is 0.25, the stripe-like non-uniform noise in the infrared image can be significantly eliminated through 3-5 iterations.
Step 107: and scanning the filtering image according to the set pixels, and gradually intercepting sub-images which take the current pixel as an origin point at the upper left corner and have the same window scale as the target image.
Step 108: and calculating a structure descriptor of the subgraph.
Step 109: calculating the similarity between the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph, specifically comprising:
using a formula
And calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph.
Where ρ isiThe cosine similarity of two vectors determined by the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph, for the structure descriptor matrix K with Y columns and each column containing X elements, the structure vector K containing X X Y elements is constructed in a way that the adjacent column vectors are connected end to end and in the sequence of the column vectors, for this reason, the cosine similarity of the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph is defined as follows:
wherein (·)' represents the transposition operation of the vector, | | · | | | represents the vector modulo operation, κTStructural descriptor K for passing differential high-frequency foregroundTStructural vector of construction, κNiFor structure descriptor K by subgraphNiThe constructed structure vector.
Step 110: and judging whether the similarity is smaller than a similarity threshold value.
Step 111: and if the similarity is smaller than a similarity threshold value, setting the corresponding area of the foreground template and the sub-image as 1.
Step 112: and judging whether the filtering image is traversed or not.
And if the similarity is greater than or equal to a similarity threshold, directly judging whether the filtering image is traversed or not.
Step 113: and if the filtering image is traversed, performing AND operation on the foreground template and the infrared source differential high-frequency foreground image to obtain a high-frequency foreground image containing a target image.
Step 114: and carrying out lossless compression on the high-frequency foreground image containing the target image to obtain a foreground compressed code stream.
Step 115: and merging the foreground compressed code stream and the background compressed code stream to obtain a mixed compressed code stream.
A simple organization mode of the single-frame infrared image mixed compressed code stream is as follows:
the hybrid compression process ends.
If not, return to step 107.
After receiving the mixed compressed code stream of the single frame image at the receiving end, the background compressed code stream and the foreground compressed code stream are respectively obtained according to the mixed compressed code stream structure given in step 114. And decompressing the background compressed code stream by adopting a corresponding lossy decoding algorithm to obtain a lossy background, decompressing the foreground compressed code stream by adopting a corresponding lossless decoding algorithm to obtain a lossless foreground, and performing addition operation on the lossy background and the lossless foreground to obtain a restored image. Fig. 2 is a schematic diagram of a hybrid compression encoding and decompression process.
The method comprises the steps of firstly realizing the separation and compression of the low-frequency background of the infrared source image through a lossy compression algorithm, and simultaneously obtaining the high-frequency foreground through differential operation with the source image and carrying out non-uniform filtering. And simultaneously, processing the target image to be retrieved by adopting the same process, and obtaining a structural descriptor of the high-frequency foreground of the target image. And then scanning the high-frequency foreground image of the source image through a sliding window, calculating a structure descriptor of a sub-image in the sliding window, simultaneously calculating the similarity between the sub-image descriptor and a target image descriptor, and synchronously updating the template when the similarity is smaller than a given threshold value. And after the high-frequency foreground scanning of the source image is finished, all the areas which do not contain the target image are '0' through the AND operation with the template. And carrying out lossless compression on the processed high-frequency foreground image to obtain a lossless compressed code stream, merging the lossless compressed code stream in the background to finally obtain a mixed compressed code stream of the source image.
Corresponding to a single-frame infrared image hybrid compression method of the present invention, the present invention further provides a single-frame infrared image hybrid compression system, as shown in fig. 3, the system includes:
the image acquisition module 201 is configured to acquire an infrared source image and a target image, and initialize a foreground template having the same size as the infrared source image.
The first lossy compression module 202 is configured to perform lossy compression on the target image, so as to obtain a target restored image.
And the structure descriptor determining module 203 of the target difference high-frequency foreground image is used for subtracting the target recovery image from the target image to obtain a target difference high-frequency foreground image and calculating a structure descriptor of the target difference high-frequency foreground.
The structural descriptor of the target differential high-frequency foreground adopts a local steering kernel, and the calculation formula of the local steering kernel is as follows:
wherein p is2The number of pixels contained in the sub-picture or the target image; h is a global smoothing parameter; matrix CiIs a covariance matrix of the spatial gradient vectors; x is the number ofjIs the coordinate of the center pixel point of the sub-image or the target image; x is the number ofiThe coordinates of other pixel points in the sub-image or the target image.
And the second lossy compression module 204 is configured to perform lossy compression on the infrared source image to obtain a background compressed code stream and a restored image of the background compressed code stream, where the restored image of the background compressed code stream is an infrared source restored image.
And the infrared source difference high-frequency foreground image determining module 205 is configured to subtract the infrared source recovery image and the infrared source image to obtain an infrared source difference high-frequency foreground image.
And the filtered image determining module 206 is configured to perform filtering processing on the infrared source difference high-frequency foreground image to obtain a filtered image.
And a sub-image determining module 207, configured to scan the filtered image according to a set pixel, and successively intercept a sub-image with a current pixel as an origin at an upper left corner and a window size the same as that of the target image.
A structure descriptor determination module 208 for a sub-graph, configured to compute a structure descriptor for the sub-graph.
And the similarity calculation module 209 is configured to calculate a similarity between the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the sub-graph.
The first determining module 210 is configured to determine whether the similarity is smaller than a similarity threshold.
And the region setting module 211 is configured to set a region corresponding to the foreground template and the sub-graph to 1 when the similarity is smaller than the similarity threshold.
A second determining module 212, configured to determine whether the filtering image is completely traversed after setting the region corresponding to the foreground template and the sub-image to 1, or determine whether the filtering image is completely traversed when the similarity is greater than or equal to the similarity threshold.
And a high-frequency foreground image determining module 213, configured to perform an and operation on the foreground template and the infrared source differential high-frequency foreground image to obtain a high-frequency foreground image including the target image when the filtered image is traversed.
And the lossless compression module 214 is configured to perform lossless compression on the high-frequency foreground image including the target image, so as to obtain a foreground compressed code stream.
And a mixed compressed code stream determining module 215, configured to combine the foreground compressed code stream and the background compressed code stream to obtain a mixed compressed code stream.
A returning module 216, configured to return to the subgraph determining module when the filtering image is not traversed.
The filtered image determining module 206 specifically includes:
the filtered image determining unit is used for filtering the infrared source difference high-frequency foreground image by adopting an anisotropic diffusion filtering method to obtain a filtered image, wherein the anisotropic diffusion filtering method adopts an iterative equation for filtering, and the iterative equation is as follows:
wherein ItIs the image at time t, It+1Is an image at the time of t +1 after one iteration; λ is a smoothing coefficient;
the divergence of the pixel (x, y) point in the four directions of east, west, south and north at the time t is respectively expressed as follows:
Sx，y、Ex，y、Nx，y、Wx，ythe heat conductivity coefficients in four directions are respectively represented, and the specific calculation formula is as follows:
wherein k is a heat conduction correlation coefficient, and the larger the value of k is, the less easy the edge is kept in the iteration process.
The similarity calculation module 209 specifically includes:
a similarity calculation unit for employing a formula
And calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph.
Where ρ isiThe cosine similarity of two vectors determined by the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph, for the structure descriptor matrix K with Y columns and each column containing X elements, the structure vector K containing X X Y elements is constructed in a way that the adjacent column vectors are connected end to end and in the sequence of the column vectors, for this reason, the cosine similarity of the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph is defined as follows:
wherein (·)' represents the transposition operation of the vector, | | · | | | represents the vector modulo operation, κTStructural descriptor K for passing differential high-frequency foregroundTStructural vector of construction, κNiFor structure descriptor K by subgraphNiThe constructed structure vector.
Example 1:
the invention provides a single-frame infrared image compression method based on the combination of lossy and lossless compression technologies, which comprises the following steps:
1. and (3) starting mixed compression, inputting an infrared source image O and a target image T, initializing a foreground template M with the same size as O to be all '0', and setting a similarity threshold V.
2. And performing lossy compression and recovery on the target image T by adopting a lossy compression algorithm such as JPEG2000, and subtracting the recovered image from the T to obtain the differential high-frequency foreground of the T. Calculating the structure descriptor K of the difference high-frequency foregroundT。
3. And (3) for the infrared source image O, performing lossy compression by adopting a lossy compression algorithm the same as the step (2) to obtain a background compressed code stream B, and subtracting the restored image of the background compressed code stream B from the restored image of the background compressed code stream B to obtain a high-frequency foreground image HF.
4. And filtering the HF to eliminate non-uniform noise and high-frequency noise of the HF to obtain a filtered image N.
5. Scanning N according to pixels, and gradually intercepting a subgraph N which takes the current pixel as an origin at the upper left corner and has the same window size as Ti。
6. Calculating N by the same method as the step (2)iStructural descriptor K ofNi。
7. Calculating KNiAnd KTSimilarity of (2)i：
Where ρ isiIs composed of KNiAnd KTCosine similarity of the two determined vectors. For a structure descriptor matrix K with Y columns and each column containing X elements, structure vectors containing X multiplied by Y elements are constructed in a way that adjacent column vectors are connected end to end and in a column vector sequence
For this purpose, KNiAnd KTCosine similarity of (c) is defined as follows:
where (·)' represents the transpose operation of the vector and | · | | | represents the vector modulo operation.
And
are respectively a pass descriptor KTAnd KNiThe constructed structure vector.
8. Judgment SiIf it is less than a given threshold value V, and if it is less than it, then the M is compared with NiAll elements of the corresponding region are set to be 1; and if the traversal of the N is not completed, carrying out the steps from fifthly to extensions.
9. And after the N is traversed, performing AND operation on the HF and the M to obtain a high-frequency foreground F containing the target T.
10. Performing lossless compression on the F, and obtaining a foreground compressed code stream by adopting a JPEG-LS lossless compression algorithm, for example; and merging the background compressed code stream B and the foreground compressed code stream into a mixed compressed code stream. The hybrid compression process ends.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (6)
1. A single-frame infrared image hybrid compression method is characterized by comprising the following steps:
acquiring an infrared source image and a target image, and initializing a foreground template with the same size as the infrared source image to be all '0';
lossy compression and recovery are carried out on the target image to obtain a target recovery image;
subtracting the target restored image from the target image to obtain a target difference high-frequency foreground image, and calculating a structure descriptor of the target difference high-frequency foreground;
lossy compression and recovery are carried out on the infrared source image to obtain a background compressed code stream and a recovered image of the background compressed code stream, wherein the recovered image of the background compressed code stream is an infrared source recovered image;
subtracting the infrared source recovery image from the infrared source image to obtain an infrared source difference high-frequency foreground image;
filtering the infrared source difference high-frequency foreground image to obtain a filtered image;
scanning the filtering image according to a set pixel, and gradually intercepting a sub-image which takes a current pixel as an origin point at the upper left corner and has the same window scale as the target image;
calculating a structure descriptor of the subgraph;
calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph; the calculating the similarity between the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph specifically comprises:
using a formula
Calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph;
where ρ isiThe cosine similarity of two vectors determined by the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph, for the structure descriptor matrix K with Y columns and each column containing X elements, the structure vector kappa containing X Y elements is constructed in the way that the adjacent column vectors are connected end to end and in the sequence of the column vectors, therefore, the structure descriptor and the sub-descriptor of the differential high-frequency foregroundThe cosine similarity of the structure descriptor of the graph is defined as follows:
wherein (·)' represents the transposition operation of the vector, | | · | | | represents the vector modulo operation, κTStructural descriptor K for passing differential high-frequency foregroundTStructural vector of construction, κNiFor structure descriptor K by subgraphNiA constructed structure vector;
judging whether the similarity is smaller than a similarity threshold value;
if the similarity is smaller than a similarity threshold value, setting the corresponding area of the foreground template and the sub-image as 1;
judging whether the filtering image is traversed or not;
if the similarity is larger than or equal to a similarity threshold, directly judging whether the filtering image is traversed or not;
if the filtering image is traversed, performing AND operation on the foreground template and the infrared source differential high-frequency foreground image to obtain a high-frequency foreground image containing a target image;
lossless compression is carried out on the high-frequency foreground image containing the target image, and a foreground compression code stream is obtained;
merging the foreground compressed code stream and the background compressed code stream to obtain a mixed compressed code stream;
and if the filtering image is not traversed, returning to scanning the filtering image according to the set pixels, and gradually intercepting sub-images which take the current pixel as an original point at the upper left corner and have the same window size as the target image.
2. The method of claim 1, wherein the structural descriptor of the target differential high-frequency foreground is a local steering kernel, and the calculation formula of the local steering kernel is as follows:
wherein p is2The number of pixels contained in the sub-picture or the target image; h is a global smoothing parameter; matrix CiIs a covariance matrix of the spatial gradient vectors; x is the number ofjIs the coordinate of the center pixel point of the sub-image or the target image; x is the number ofiThe coordinates of other pixel points in the sub-image or the target image.
3. The method for compressing the single-frame infrared image in a mixed manner according to claim 1, wherein the filtering of the infrared source difference high-frequency foreground image to obtain a filtered image specifically comprises:
filtering the infrared source difference high-frequency foreground image by adopting an anisotropic diffusion filtering method to obtain a filtered image, wherein the anisotropic diffusion filtering method adopts an iterative equation for filtering, and the iterative equation is as follows:
wherein ItIs the image at time t, It+1Is an image at the time of t +1 after one iteration; λ is a smoothing coefficient;
the divergence of the pixel (x, y) point in the four directions of east, west, south and north at the time t is respectively expressed as follows:
Sx,y、Ex,y、Nx,y、Wx,ythe heat conductivity coefficients in four directions are respectively represented, and the specific calculation formula is as follows:
wherein k is a heat conduction correlation coefficient, and the larger the value of k is, the less easy the edge is kept in the iteration process.
4. A single frame infrared image hybrid compression system, comprising:
the image acquisition module is used for acquiring an infrared source image and a target image and initializing a foreground template with the same size as the infrared source image to be all '0';
the first lossy compression module is used for carrying out lossy compression and recovery on the target image to obtain a target recovery image;
a structure descriptor determining module of the target difference high-frequency foreground image, which is used for subtracting the target recovery image from the target image to obtain a target difference high-frequency foreground image and calculating a structure descriptor of the target difference high-frequency foreground;
the second lossy compression module is used for performing lossy compression and recovery on the infrared source image to obtain a background compressed code stream and a recovered image of the background compressed code stream, wherein the recovered image of the background compressed code stream is an infrared source recovered image;
the infrared source difference high-frequency foreground image determining module is used for subtracting the infrared source recovery image from the infrared source image to obtain an infrared source difference high-frequency foreground image;
the filtering image determining module is used for carrying out filtering processing on the infrared source difference high-frequency foreground image to obtain a filtering image;
a subgraph determining module, which is used for scanning the filtering image according to the set pixel and gradually intercepting the subgraph which takes the current pixel as the origin of the upper left corner and has the same window scale as the target image;
the structure descriptor determining module of the subgraph is used for calculating the structure descriptor of the subgraph;
the similarity calculation module is used for calculating the similarity between the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph; the similarity calculation module specifically includes:
a similarity calculation unit for employing a formula
Calculating the similarity of the structure descriptor of the target differential high-frequency foreground and the structure descriptor of the subgraph;
where ρ isiThe cosine similarity of two vectors determined by the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph, for the structure descriptor matrix K with Y columns and each column containing X elements, the structure vector K containing X X Y elements is constructed in a way that the adjacent column vectors are connected end to end and in the sequence of the column vectors, for this reason, the cosine similarity of the structure descriptor of the differential high-frequency foreground and the structure descriptor of the subgraph is defined as follows:
wherein (·)' represents the transposition operation of the vector, | | · | | | represents the vector modulo operation, κTStructural descriptor K for passing differential high-frequency foregroundTStructural vector of construction, κNiFor structure descriptor K by subgraphNiA constructed structure vector;
the first judgment module is used for judging whether the similarity is smaller than a similarity threshold value;
the region setting module is used for setting the region corresponding to the foreground template and the sub-image as 1 when the similarity is smaller than the similarity threshold;
the second judging module is used for judging whether the filtering image is completely traversed after the area corresponding to the foreground template and the sub-image is set to be 1 or judging whether the filtering image is completely traversed when the similarity is larger than or equal to the similarity threshold;
the high-frequency foreground image determining module is used for performing AND operation on the foreground template and the infrared source differential high-frequency foreground image to obtain a high-frequency foreground image containing a target image when the filter image is traversed;
the lossless compression module is used for carrying out lossless compression on the high-frequency foreground image containing the target image to obtain a foreground compressed code stream;
a mixed compressed code stream determining module, configured to combine the foreground compressed code stream and the background compressed code stream to obtain a mixed compressed code stream;
and the return module is used for returning to the subgraph determination module when the filtering image is not traversed.
5. The single-frame infrared image hybrid compression system of claim 4, wherein the structural descriptor of the target differential high-frequency foreground adopts a local steering kernel, and a calculation formula of the local steering kernel is as follows:
wherein p is2The number of pixels contained in the sub-picture or the target image; h is a global smoothing parameter; matrix CiIs a covariance matrix of the spatial gradient vectors; x is the number ofjIs the coordinate of the center pixel point of the sub-image or the target image; x is the number ofiThe coordinates of other pixel points in the sub-image or the target image.
6. The single-frame infrared image hybrid compression system according to claim 4, wherein the filtered image determination module specifically includes:
the filtered image determining unit is used for filtering the infrared source difference high-frequency foreground image by adopting an anisotropic diffusion filtering method to obtain a filtered image, wherein the anisotropic diffusion filtering method adopts an iterative equation for filtering, and the iterative equation is as follows:
wherein ItIs the image at time t, It+1Is an image at the time of t +1 after one iteration; λ is a smoothing coefficient;
the divergence of the pixel (x, y) point in the four directions of east, west, south and north at the time t is respectively expressed as follows:
Sx,y、Ex,y、Nx,y、Wx,ythe heat conductivity coefficients in four directions are respectively represented, and the specific calculation formula is as follows:
wherein k is a heat conduction correlation coefficient, and the larger the value of k is, the less easy the edge is kept in the iteration process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011088307.3A CN112399182B (en) | 2020-10-13 | 2020-10-13 | Single-frame infrared image hybrid compression method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011088307.3A CN112399182B (en) | 2020-10-13 | 2020-10-13 | Single-frame infrared image hybrid compression method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112399182A CN112399182A (en) | 2021-02-23 |
| CN112399182B true CN112399182B (en) | 2021-08-31 |
Family
ID=74595523
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011088307.3A Active CN112399182B (en) | 2020-10-13 | 2020-10-13 | Single-frame infrared image hybrid compression method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112399182B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102792245A (en) * | 2010-03-16 | 2012-11-21 | 英特飞斯公司 | Interactive display system |
| CN103297754A (en) * | 2013-05-02 | 2013-09-11 | 上海交通大学 | Monitoring video self-adaption interesting area coding system |
| CN110826554A (en) * | 2018-08-10 | 2020-02-21 | 西安电子科技大学 | Infrared target detection method |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8238678B2 (en) * | 2006-08-30 | 2012-08-07 | Siemens Medical Solutions Usa, Inc. | Providing representative image information |
| US9131913B2 (en) * | 2012-06-14 | 2015-09-15 | Carestream Health, Inc. | Region-selective fluoroscopic image compression |
| CN106663316A (en) * | 2016-08-30 | 2017-05-10 | 深圳大学 | Block sparse compressive sensing-based infrared image reconstruction method and system thereof |
| CN106683056A (en) * | 2016-12-16 | 2017-05-17 | 凯迈（洛阳）测控有限公司 | Airborne photoelectric infrared digital image processing method and apparatus thereof |
-
2020
- 2020-10-13 CN CN202011088307.3A patent/CN112399182B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102792245A (en) * | 2010-03-16 | 2012-11-21 | 英特飞斯公司 | Interactive display system |
| CN103297754A (en) * | 2013-05-02 | 2013-09-11 | 上海交通大学 | Monitoring video self-adaption interesting area coding system |
| CN110826554A (en) * | 2018-08-10 | 2020-02-21 | 西安电子科技大学 | Infrared target detection method |
Non-Patent Citations (2)
| Title |
|---|
| Temporal and spatial compression of infrared imagery sequences containing slow motion point target;Revital Huber-Shalem; Ofer Hadar; Stanley R. Rotman; et al;《2012 IEEE 27th Convention of Electrical and Electronics Engineers in Isreal》;20121117;全文 * |
| 红外小目标实时图像实时压缩传输系统的研究;张锦龙;《中国优秀硕士学位论文全文数据库（电子期刊）》;20170515;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112399182A (en) | 2021-02-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Image compressive sensing recovery via collaborative sparsity | |
| CN102123278B (en) | Signal source encoding and decoding method based on distributed compressive sensing technology | |
| US8345971B2 (en) | Method and system for spatial-temporal denoising and demosaicking for noisy color filter array videos | |
| CN101980537B (en) | Object and fractal-based binocular three-dimensional video compression coding and decoding method | |
| CN110933429B (en) | Video compression sensing and reconstruction method and device based on deep neural network | |
| CN104754340B (en) | Unmanned aerial vehicle reconnaissance image compression method | |
| Shirani et al. | Reconstruction of baseline JPEG coded images in error prone environments | |
| CN102291579B (en) | Rapid fractal compression and decompression method for multi-cast stereo video | |
| CN110870310A (en) | Image encoding method and apparatus | |
| US9268791B2 (en) | Method and apparatus for image processing and computer readable medium | |
| CN102316323B (en) | Rapid binocular stereo-video fractal compressing and uncompressing method | |
| CN112991450A (en) | Detail enhancement unsupervised depth estimation method based on wavelet | |
| Yuan et al. | Object shape approximation and contour adaptive depth image coding for virtual view synthesis | |
| CN114651270A (en) | Depth loop filtering by time-deformable convolution | |
| CN112399182B (en) | Single-frame infrared image hybrid compression method and system | |
| Sun et al. | A Novel adaptive directional interpolation algorithm for digital video resolution enhancement | |
| Coluccia et al. | Smoothness-constrained image recovery from block-based random projections | |
| CN107509074B (en) | Self-adaptive 3D video compression coding and decoding method based on compressed sensing | |
| CN102263952B (en) | Quick fractal compression and decompression method for binocular stereo video based on object | |
| CN102263954B (en) | Quick fractal video compression and decompression method based on object | |
| CN102263953B (en) | Quick fractal compression and decompression method for multicasting stereo video based on object | |
| Abrardo et al. | A compressive sampling scheme for iterative hyperspectral image reconstruction | |
| CN103647969A (en) | Object-based fast fractal video compression and decompression method | |
| Zhu et al. | JPEG lifting algorithm based on adaptive block compressed sensing | |
| Fang et al. | Priors guided extreme underwater image compression for machine vision and human vision |
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 |