CN110599466A - Hyperspectral anomaly detection method for component projection optimization separation - Google Patents
Hyperspectral anomaly detection method for component projection optimization separation Download PDFInfo
- Publication number
- CN110599466A CN110599466A CN201910807948.0A CN201910807948A CN110599466A CN 110599466 A CN110599466 A CN 110599466A CN 201910807948 A CN201910807948 A CN 201910807948A CN 110599466 A CN110599466 A CN 110599466A
- Authority
- CN
- China
- Prior art keywords
- pixel
- super
- component projection
- calculating
- hyperspectral
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a hyperspectral anomaly detection method for component projection optimization separation, which comprises the steps of performing superpixel segmentation on a hyperspectral image by using a superpixel segmentation algorithm based on entropy rate; calculating the average value of pixel points contained in each super pixel to be used as the spectral vector of the super pixel; calculating the mahalanobis distance from the pixel point contained in each super pixel to the spectrum vector, and summing the discrete values representing the super pixels; taking the super-pixels as units, and solving local abnormal factors one by one; calculating the product of the discrete value of each super pixel and the reciprocal of the local abnormal factor thereof, and selecting partial super pixels with smaller products as a background set to construct a pre-estimated background set; setting a component projection and separation optimization filter function, and solving an optimal filter vector; and multiplying the optimal solution and the hyperspectral image pixel by pixel to obtain a detection result. The method comprises the steps of reading image information from a super-pixel level, acquiring a pre-estimated background set by using local abnormal factors, and obtaining a hyperspectral image abnormal target detection result by combining optimized filtering.
Description
Technical Field
The invention belongs to the technical field of computer image processing, relates to an image target abnormity detection method, and particularly relates to a hyperspectral abnormity detection method based on component projection optimization separation.
Background
The hyperspectral image comprises a plurality of wave bands, and compared with a gray level image and an RGB image, the hyperspectral image contains more spectral information, so that abnormal ground objects which are obviously different from spectral information of surrounding ground objects can be detected by utilizing the difference of spectral characteristic curves of different ground objects. Although the hyperspectral images are increasingly used with the maturity of the hyperspectral imaging technology and the reduction of the cost, some limiting conditions still exist in the anomaly detection technology for the hyperspectral images.
1) The hyperspectral image has very fine spectral resolution, but the spatial resolution is generally low, so that the target ground object with detection often exists in the image in the form of extremely small pixels and even sub-pixels.
2) The traditional detection method based on the model still has the problem of insufficient background suppression.
3) Due to the fact that no prior information exists and the phenomena of 'same object, different spectrum, and same foreign object spectrum' possibly exist in the hyperspectral image, some algorithms cannot correctly detect the interested target by processing the hyperspectral image, namely the prominent effect on target information is not obvious enough.
Therefore, a detection method capable of sufficiently suppressing the response information of the ground object in the background of the hyperspectral image and highlighting the response value of the interested abnormal target to be detected is needed. The method in "A Background-analysis Component project and Separation Optimized Filter for analysis Detection in Hyperspectral images" still has the situation of wrong classification of potential Background and abnormal classes in practical application, and needs to be further improved.
Disclosure of Invention
In order to solve the technical problems, the invention provides a hyperspectral anomaly detection method based on the prior art, wherein the hyperspectral anomaly detection method is used for performing component projection optimization separation, and is used for performing superpixel segmentation processing on a data set and detecting a potential anomaly target at the superpixel level.
The invention provides a hyperspectral anomaly detection method for component projection optimization separation, which comprises the following steps of:
step 1, performing superpixel segmentation on a hyperspectral image by using a superpixel segmentation algorithm based on entropy rate;
step 2, calculating the average value of pixel points contained in each super pixel to be used as the spectral vector of the super pixel;
step 3, calculating the Mahalanobis distance from the pixel point contained in each super pixel to the spectrum vector, and summing the discrete values representing the super pixels;
step 4, with the super-pixel as a unit, solving local abnormal factors one by one;
step 5, calculating the product of the discrete value of each super pixel and the reciprocal of the local abnormal factor thereof, and selecting partial super pixels with smaller products as a background set to construct a pre-estimated background set;
step 6, setting a component projection and separation optimization filter function;
step 7, solving the optimal filter vector of the component projection and separation optimization filter function;
and 8, multiplying the optimal solution and the hyperspectral image pixel by pixel to obtain a hyperspectral anomaly detection result.
Furthermore, in step 4, the local anomaly factor is calculated for the superpixel, and the implementation process includes the following substeps,
step 4.1, calculating the k-adjacent distance of the current super pixel p;
step 4.2, calculating the reachable distance from any super pixel in the k-neighborhood to the current super pixel p;
4.3, calculating the local reachable density of the current superpixel p in a way of taking the reciprocal of the average reachable distance from the superpixel in the k-neighborhood of p to p;
and 4.4, obtaining the local outlier factor of the current superpixel p, wherein the calculation mode is to take the average of the ratio of the local reachable density of the superpixel in the k-neighborhood of p to the local reachable density of p.
Furthermore, the construction of the pre-estimated background set in step 5 is realized by calculating the product of the discrete value of each super pixel and the reciprocal of the corresponding local abnormal factor, and selecting 80% of super pixels with smaller products as the background set.
Furthermore, the setting of the component projection and separation optimization filter function in step 6, the implementation process includes the following sub-steps,
step 6.1, calculate the average component projection f of the image1，
f1＝wTΛ-1VTCVΛ-1w
C is a covariance matrix of the hyperspectral image, V and Λ are respectively corresponding to a matrix formed by eigenvectors of the covariance matrix C and a diagonal matrix formed by corresponding eigenvalues of the covariance matrix C, and w is an optimal filter vector to be solved;
step 6.2, calculating the product f of the pre-estimated background set and the weighted value thereof2，
f2＝wTXB
Wherein, XBIs the estimated background set obtained in step 5;
step 6.3, projecting the average component f by a Lagrange multiplier1And the product f of the estimated background set and the weighted value thereof2To obtain a component projection and separation optimization filter function:
where λ is the lagrange multiplier.
In step 7, the solution is performed by using the FISTA algorithm to obtain the optimal filter vector of the component projection and separation optimization filter function.
The method comprises the steps of reading image information from a superpixel layer, acquiring an estimated background set by using local abnormal factors, solving average component projection of the image by combining an optimization filtering theory, designing component projection and separating and optimizing a filtering function, so that the salient of abnormal components and the inhibition of background components in hyperspectral data are realized, and a final result of hyperspectral image abnormal target detection is obtained. Compared with the prior art, the invention has the beneficial effects that:
(1) the invention innovatively interprets and processes the hyperspectral data at a superpixel level, and can effectively avoid the situation that the distance of pixel points in a high-dimensional space is difficult to distinguish due to excessive hyperspectral dimensions; introducing local abnormal factors at the superpixel level to express the distribution density of the current superpixel compared with the neighborhood points of the current superpixel, and further obtaining a more stable potential background set; the anomaly detection according to the present invention can sufficiently suppress the detection output energy of the background point by processing the estimated background set.
(2) The invention designs a new component projection and separation optimization filter function: by combining an optimization filtering theory, the average component projection is solved by fully utilizing the statistical distribution characteristics of the image, so that the abnormal point to be detected has larger output energy, meanwhile, the response value of the pre-estimated background set is minimized, the superiority of a super pixel layer is considered on the basis that the potential background category has smaller output energy, and the participation in the aspect of pre-estimated weight values corresponding to pixels in the background set is omitted. And then, using a FISTA algorithm to optimize and solve the component projection and separation functions so as to obtain the optimal filter vector.
(3) The invention realizes more stable suppression of background information and reasonably detects the abnormal target to be detected without prior.
Drawings
Fig. 1 is a schematic flow chart of an embodiment of the present invention.
FIG. 2 shows a local anomaly factor solution procedure according to an embodiment of the present invention.
Detailed Description
In order to facilitate the understanding and implementation of the present invention for those of ordinary skill in the art, the present invention is further described in detail with reference to the accompanying drawings and examples, it is to be understood that the embodiments described herein are merely illustrative and explanatory of the present invention and are not restrictive thereof.
Referring to fig. 1, the hyperspectral anomaly detection method for component projection optimization separation provided by the embodiment of the invention includes the following steps:
step 1: performing Superpixel Segmentation on the hyperspectral image by using an Entropy Rate-based Superpixel Segmentation algorithm (Entropy Rate Superpixel Segmentation), and generating n superpixels;
the super-pixel segmentation algorithm based on the entropy rate is the prior art, and is not repeated in the invention; in specific implementation, the value of n can be preset by those skilled in the art.
Step 2: and calculating the average value of pixel points contained in each super pixel to be used as the spectral vector of the super pixel.
And step 3: and calculating the mahalanobis distance from the pixel point contained in each super pixel to the spectrum vector, and summing the discrete values representing the super pixels.
And 4, step 4: taking the super-pixels as units, and solving local abnormal factors one by one;
referring to fig. 2, a current superpixel of a hyperspectral image is represented by p, a k-neighborhood distance of p is calculated first, then a local reachable density of the superpixel located within the k-neighborhood distance is calculated, and finally a local abnormal factor of p is obtained.
In an embodiment, the specific implementation process of solving a local abnormal factor for any super pixel includes the following sub-steps:
step 4.1: the k-neighborhood distance of the current superpixel p is calculated. Firstly, calculating Euclidean distances from a spectrum vector of p to other superpixel spectrum vectors, sequencing the Euclidean distances from small to large, and selecting k superpixels with the smallest Euclidean distances as a k-distance neighborhood of p, wherein the k-large distance is the k-adjacent distance of p; in specific implementation, the value of k can be preset by a person skilled in the art;
step 4.2: calculating the reachable distance from any super pixel in the k-neighborhood to the current super pixel p: the reachable distance of the superpixel o to the current superpixel p is calculated. When the Euclidean distance from o to p is less than or equal to the k-adjacent distance of p, the reachable distance is the k-adjacent distance of p; when the Euclidean distance from o to p is larger than the k-adjacent distance of p, the reachable distance is the Euclidean distance between two superpixels;
step 4.3: calculating the local reachable density of the current superpixel p in a way of taking the reciprocal of the average reachable distance from the superpixel in the k-neighborhood of p to p;
step 4.4: the local outlier factor for the current superpixel p is obtained by taking the average of the ratio of the local reachable density of superpixels in the k-neighborhood of p to the local reachable density of p.
And 5: calculating the product of the discrete value of each super pixel of the hyperspectral image and the reciprocal of the local abnormal factor of the super pixel, and selecting partial super pixels with smaller products as a background set to construct a pre-estimated background set;
in an embodiment, the construction of the predictive background set is realized by calculating a product of a discrete value of each super pixel and an inverse of a local abnormal factor thereof, and selecting 80% of super pixels with smaller products as the background set.
Step 6: setting a component projection and separation optimization filter function;
in an embodiment, the step 6 specifically includes the following sub-steps:
step 6.1: calculating the mean component projection f of the image1：
f1＝wTΛ-1VTCVΛ-1w
Wherein C is a covariance matrix of the hyperspectral image, V and Λ are respectively corresponding to a matrix formed by eigenvectors of the covariance matrix C and a diagonal matrix formed by corresponding eigenvalues of the covariance matrix C, and w is an optimal filter vector to be solved by the method.
Step 6.2: calculating the product f of the pre-estimated background set and the weighted value thereof2：
f2＝wTXB
Wherein XBAnd 5, selecting the obtained pre-estimated background set through local abnormal factors to obtain a super-pixel background set.
Step 6.3: projecting the average component f by a Lagrange multiplier1And the product f of the estimated background set and the weighted value thereof2To obtain a component projection and separation optimization filter function f:
where λ is the lagrange multiplier.
And 7: solving the optimal filter vector of the component projection and separation optimization filter function by using an optimization algorithm;
in an embodiment, the optimal filter vector of the component projection and the separation optimization filter function is solved by using an optimization algorithm, preferably, the optimal filter vector is solved by using a FISTA algorithm, and the FISTA (a fast iterative threshold-threshold shrinkage algorithm) is a fast iterative threshold shrinkage algorithm and is not described herein again for the prior art.
And 8: multiplying the optimal filtering vector obtained in the step 7 with the hyperspectral image pixel by pixel to obtain a hyperspectral anomaly detection result;
the steps for realizing the hyperspectral image abnormal target detection method are as follows. Through the introduction of super-pixel segmentation, component projection, local abnormal factors and optimized filtering, the advantages of a hyperspectral image can be fully utilized, and the abnormal target is highlighted while background energy is suppressed. In specific implementation, the automatic operation of the process can be realized by adopting a software mode. The apparatus for operating the process should also be within the scope of the present invention.
The embodiment is realized by adopting an MATLAB platform, and an MATLAB hyperspectral remote sensing image read-write function is taken as an implementation basis. Calling a hyperspectral remote sensing image reading function, inputting a file name of a remote sensing image to be read, reading the remote sensing image into a matrix with the size of N x P, wherein each element in the matrix is a pixel radiation value corresponding to each wave band, N is the pixel number of the remote sensing image, and P is the wave band number of the remote sensing image. The MATLAB remote sensing image read-write function is a well-known technology in the art and is not described herein.
It should be understood that parts of the specification not set forth in detail are well within the prior art.
It should be understood that the above description of the preferred embodiments is given for clarity and not for any purpose of limitation, and that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (6)
1. A hyperspectral anomaly detection method for component projection optimization separation is characterized by comprising the following steps:
step 1, performing superpixel segmentation on a hyperspectral image by using a superpixel segmentation algorithm based on entropy rate;
step 2, calculating the average value of pixel points contained in each super pixel to be used as the spectral vector of the super pixel;
step 3, calculating the Mahalanobis distance from the pixel point contained in each super pixel to the spectrum vector, and summing the discrete values representing the super pixels;
step 4, with the super-pixel as a unit, solving local abnormal factors one by one;
step 5, calculating the product of the discrete value of each super pixel and the reciprocal of the local abnormal factor thereof, and selecting partial super pixels with smaller products as a background set to construct a pre-estimated background set;
step 6, setting a component projection and separation optimization filter function;
step 7, solving the optimal filter vector of the component projection and separation optimization filter function;
and 8, multiplying the optimal solution and the hyperspectral image pixel by pixel to obtain a hyperspectral anomaly detection result.
2. The hyperspectral anomaly detection method by component projection optimization separation according to claim 1, characterized in that: in step 4, local abnormal factors are obtained for the superpixels, and the implementation process comprises the following substeps,
step 4.1, calculating the k-adjacent distance of the current super pixel p;
step 4.2, calculating the reachable distance from any super pixel in the k-neighborhood to the current super pixel p;
4.3, calculating the local reachable density of the current superpixel p in a way of taking the reciprocal of the average reachable distance from the superpixel in the k-neighborhood of p to p;
and 4.4, obtaining the local outlier factor of the current superpixel p, wherein the calculation mode is to take the average of the ratio of the local reachable density of the superpixel in the k-neighborhood of p to the local reachable density of p.
3. The hyperspectral anomaly detection method by component projection optimization separation according to claim 1, characterized in that: and 5, constructing the pre-estimated background set, wherein the construction process is realized by calculating the product of the discrete value of each super pixel and the reciprocal of the corresponding local abnormal factor, and 80% of super pixels with smaller products are selected as the background set.
4. The hyperspectral anomaly detection method by component projection optimization separation according to claim 1, 2 or 3, characterized in that: the component projection and separation optimization filter function is set in the step 6, and the implementation process comprises the following sub-steps,
step 6.1, calculate the average component projection f of the image1，
f1＝wTΛ-1VTCVΛ-1w
C is a covariance matrix of the hyperspectral image, V and Λ are respectively corresponding to a matrix formed by eigenvectors of the covariance matrix C and a diagonal matrix formed by corresponding eigenvalues of the covariance matrix C, and w is an optimal filter vector to be solved;
step 6.2, calculating the product f of the pre-estimated background set and the weighted value thereof2，
f2＝wTXB
Wherein, XBIs the estimated background set obtained in step 5;
step 6.3, projecting the average component f by a Lagrange multiplier1And the product f of the estimated background set and the weighted value thereof2To obtain a component projection and separation optimization filter function:
where λ is the lagrange multiplier.
5. The hyperspectral anomaly detection method by component projection optimization separation according to claim 1, 2 or 3, characterized in that: in step 7, solving is carried out by using a FISTA algorithm to obtain an optimal filtering vector of the component projection and separation optimization filtering function.
6. The hyperspectral anomaly detection method by component projection optimization separation according to claim 4, characterized in that: in step 7, solving is carried out by using a FISTA algorithm to obtain an optimal filtering vector of the component projection and separation optimization filtering function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910807948.0A CN110599466B (en) | 2019-08-29 | 2019-08-29 | Hyperspectral anomaly detection method for component projection optimization separation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910807948.0A CN110599466B (en) | 2019-08-29 | 2019-08-29 | Hyperspectral anomaly detection method for component projection optimization separation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110599466A true CN110599466A (en) | 2019-12-20 |
| CN110599466B CN110599466B (en) | 2022-04-29 |
Family
ID=68856312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910807948.0A Active CN110599466B (en) | 2019-08-29 | 2019-08-29 | Hyperspectral anomaly detection method for component projection optimization separation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110599466B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329791A (en) * | 2020-10-30 | 2021-02-05 | 中国电子科技集团公司第五十四研究所 | Automatic extraction method for hyperspectral image water area |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105069758A (en) * | 2015-08-21 | 2015-11-18 | 武汉大学 | Hyperspectral image denoising method based on robust low-rank tensor |
| US20160110844A1 (en) * | 2014-10-21 | 2016-04-21 | The Boeing Company | Multiple Pixel Pitch Super Resolution |
| CN105551000A (en) * | 2015-12-11 | 2016-05-04 | 中国科学院遥感与数字地球研究所 | Remote sensing image reconstruction method based on reference image structure constraint and non-convex low rank constraint |
| CN105825512A (en) * | 2016-03-18 | 2016-08-03 | 武汉大学 | Hyperspectral remote sensing image abnormal target detection method based on robust background regression |
| CN107784655A (en) * | 2016-12-28 | 2018-03-09 | 中国测绘科学研究院 | A kind of visual attention model SAR naval vessels detection algorithm of adaptive threshold |
| CN109190511A (en) * | 2018-08-14 | 2019-01-11 | 西北工业大学 | Hyperspectral classification method based on part Yu structural constraint low-rank representation |
-
2019
- 2019-08-29 CN CN201910807948.0A patent/CN110599466B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160110844A1 (en) * | 2014-10-21 | 2016-04-21 | The Boeing Company | Multiple Pixel Pitch Super Resolution |
| CN105069758A (en) * | 2015-08-21 | 2015-11-18 | 武汉大学 | Hyperspectral image denoising method based on robust low-rank tensor |
| CN105551000A (en) * | 2015-12-11 | 2016-05-04 | 中国科学院遥感与数字地球研究所 | Remote sensing image reconstruction method based on reference image structure constraint and non-convex low rank constraint |
| CN105825512A (en) * | 2016-03-18 | 2016-08-03 | 武汉大学 | Hyperspectral remote sensing image abnormal target detection method based on robust background regression |
| CN107784655A (en) * | 2016-12-28 | 2018-03-09 | 中国测绘科学研究院 | A kind of visual attention model SAR naval vessels detection algorithm of adaptive threshold |
| CN109190511A (en) * | 2018-08-14 | 2019-01-11 | 西北工业大学 | Hyperspectral classification method based on part Yu structural constraint low-rank representation |
Non-Patent Citations (5)
| Title |
|---|
| SAFA KHAZAI ET AL.: "An Approach for Subpixel Anomaly Detection in Hyperspectral Images", 《IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING》 * |
| SHIZHEN CHANG ET AL.: "BASO: A Background-Anomaly Component BASO: A Background-Anomaly Component Anomaly Detection in Hyperspectral Images", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 * |
| WEIWEI SUN ET AL.: "A Randomized Subspace Learning Based Anomaly Detector for Hyperspectral Imagery", 《REMOTE SENSING》 * |
| 段无悔: "基于超像素的高光谱图像分类算法研究", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
| 赵锐 等: "一种基于核特征空间的鲁棒性高光谱异常探测方法", 《光子学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329791A (en) * | 2020-10-30 | 2021-02-05 | 中国电子科技集团公司第五十四研究所 | Automatic extraction method for hyperspectral image water area |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110599466B (en) | 2022-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Image interpolation via regularized local linear regression | |
| US8203554B2 (en) | Method and apparatus for identifying visual content foregrounds | |
| Jiao et al. | Local stereo matching with improved matching cost and disparity refinement | |
| EP4109392A1 (en) | Image processing method and image processing device | |
| Huhle et al. | Robust non-local denoising of colored depth data | |
| Lo et al. | Edge-preserving depth map upsampling by joint trilateral filter | |
| JP6136537B2 (en) | Image processing apparatus, image processing method, image processing control program, and recording medium | |
| US9589323B1 (en) | Super resolution image enhancement technique | |
| JP2014002738A (en) | Device and method for refining and decomposing material composition of mixed pixels in remote sensing image | |
| US20160078634A1 (en) | Methods and systems for image matting and foreground estimation based on hierarchical graphs | |
| EP3073443B1 (en) | 3d saliency map | |
| CN110427997B (en) | Improved CVA change detection method for complex remote sensing image background | |
| Luo et al. | Convolutional neural networks-based stereo image reversible data hiding method | |
| CN107392211B (en) | Salient target detection method based on visual sparse cognition | |
| CN114092459A (en) | Power equipment thermal fault diagnosis method, equipment and medium based on decision-level fusion | |
| Zhang et al. | Single image dehazing based on bright channel prior model and saliency analysis strategy | |
| CN110599466B (en) | Hyperspectral anomaly detection method for component projection optimization separation | |
| JP2013182330A (en) | Image processor and image processing method | |
| Wang et al. | A unified framework of source camera identification based on features | |
| US8897378B2 (en) | Selective perceptual masking via scale separation in the spatial and temporal domains using intrinsic images for use in data compression | |
| Zhou et al. | An optimal higher order likelihood distribution based approach for strong edge and high contrast restoration | |
| Rezaei Abkenar et al. | Salient region detection using feature extraction in the non‐subsampled contourlet domain | |
| Yang et al. | Depth map reconstruction and rectification through coding parameters for mobile 3D video system | |
| Sun et al. | Image comparison by compound disjoint information with applications to perceptual visual quality assessment, image registration and tracking | |
| Qiao et al. | A novel SVM-based blind super-resolution algorithm |
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 |