CN106651781A - Image noise suppression method for laser active imaging - Google Patents

Image noise suppression method for laser active imaging Download PDF

Info

Publication number
CN106651781A
CN106651781A CN201610850521.5A CN201610850521A CN106651781A CN 106651781 A CN106651781 A CN 106651781A CN 201610850521 A CN201610850521 A CN 201610850521A CN 106651781 A CN106651781 A CN 106651781A
Authority
CN
China
Prior art keywords
image
filtering
structural element
pixel
scale
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
Application number
CN201610850521.5A
Other languages
Chinese (zh)
Other versions
CN106651781B (en
Inventor
王宇
朴燕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Changchun University of Science and Technology
Original Assignee
Changchun University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Changchun University of Science and Technology filed Critical Changchun University of Science and Technology
Priority to CN201610850521.5A priority Critical patent/CN106651781B/en
Publication of CN106651781A publication Critical patent/CN106651781A/en
Application granted granted Critical
Publication of CN106651781B publication Critical patent/CN106651781B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06T2207/10032—Satellite or aerial image; Remote sensing
    • G06T2207/10044—Radar 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
    • G06T2207/10—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
    • G06T2207/20036—Morphological image processing
    • 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
    • G06T2207/20172—Image enhancement details
    • G06T2207/20192—Edge enhancement; Edge preservation

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to an image noise suppression method for laser active imaging, and belongs to an image noise suppression method. Structural elements of different scales are used to carry out morphological opening-closing operation and morphological closing-opening operation on an image formed by laser active imaging, and the results of multi-scale morphological filtering are fused according to the homogeneity of the local areas of the image. Multi-scale morphological image de-noising refers to weighted averaging of the morphological filtering results of the structural elements of all scales. Based on a morphological filtering method, a hardware logic structure is easy to implement. The image edge can be processed in real time and maintained properly. By using structural elements of multiple scales in morphological filtering and de-noising and taking the homogeneity of the local areas of the image as the basis for the fusion of results of image filtering by structural elements of different scales, a better laser image de-noising effect can be obtained. The method can be applied to laser active imaging systems, synthetic aperture radars, infrared medical imaging and other image de-noising occasions with speckle noise.

Description

A kind of image noise suppression method of Laser active illuminated imaging
Technical field
The present invention relates to a kind of image noise suppression method, more particularly to a kind of to utilize image-region homogeneity to multiple dimensioned The laser image denoising method that morphologic filtering result is merged.
Background technology
In recent years, Laser active illuminated imaging technology is developed rapidly, because the technology can provide more stable, clearly mesh Logo image, can provide more rich information for accurate description target geometry, thus radar-reconnaissance, target acquisition with The military fields such as track, precise guidance are with a wide range of applications.But, the target image under laser irradiation can be dissipated by laser The modulation of spot, the presence of speckle noise makes the quality degradation of image, and this necessarily affects the identification to target and tracking accuracy, So research noise suppressing method of speckle noise suitable for laser image has important practical value.
Speckle noise is signal dependent noise, is substantially nonlinear, removal more difficult than additive noise.Chinese scholars are carried Various Image filter arithmetics are gone out.Classical algorithm has Lee, Kuan, SBF and wavelet filtering etc., and these methods exist While suppressing speckle noise, many edge detail informations also receive loss.Non-local mean filtering (NLM) can be fully sharp With the information contained in entire image, retain the texture structure of image well while noise is effectively suppressed, but it is performed Computation complexity is larger, and the algorithm speed of service is slower.
It is easy to hardware logic structure based on morphologic filtering method to realize, schemes with energy real-time processing and preferably holding When being used for laser image denoising as the advantage at edge, but traditional morphologic filtering algorithm, its noise compressed capability is poor, so If can be improved to traditional form filtering algorithm, it is possible on the premise of image border is kept, laser image is improved In speckle noise suppressing ability.
The content of the invention
The present invention provides a kind of image noise suppression method of Laser active illuminated imaging, with reference to laser image feature, it is proposed that A kind of multi-scale morphology filtering method, morphologic filtering process is carried out by laser image under multiple yardsticks, then fully sharp It is strong with large-scale structure element filter effect and small-scale structure element keeps edge details clearly advantage, according to the figure of calculating As the homogeneous property coefficient in region, the morphologic filtering result under different scale is merged, reaching can suppress noise energy holding figure again As the purpose at edge.
The present invention is adopted the technical scheme that, comprised the following steps：
(1), using different scale structural element, respectively form open-close and form are carried out to the image of Laser active illuminated imaging Close-open filtering operation；
" disk " structural element of different scale is chosen, radius is respectively d1、d2……dn, structural element Bd(i)Represent, Bd(1)It is d for radius1" disk " structural element, Bd(2)It is d for radius2" disk " structural element, by that analogy；If F (x, Y) it is input signal, S (x, y) is output signal, using structural element Bd(i)Make form open-close to input signal and form close- Opening operation, operation result is expressed as OCd(i)(x,y)、COd(i)(x,y)；
Form open-close and form are closed-opening operation result combines：
Wherein,Morphology opening operation is represented,Represent closing operation of mathematical morphology, OCCOd(i)(x, y) is represented and is utilized radius For di" disk " structural element carry out form open-close and form closes the-filter result image that combines of opening operation；
(2) according to the homogeneity of image local area, by the fusion of multi-scale morphology filtering result；
(1). judge whether each pixel belongs to homogenous region in image
For each image pixel, surrounding M × M neighborhoods are chosen, according to formula (8), filtering figure under each yardstick is calculated respectively As OCCOd(i)The area uniformity of (x, y), is as a result designated as Hd(i)(x,y)；
Wherein, f (x, y) represents the M × M neighborhoods centered on pixel (x, y), and Var (f (x, y)), E [f (x, y)] are respectively Represent the variance and average of grey scale pixel value in regional area；
For judging the threshold value of image-region homogeneity for Hthresh, work as Hd(i)≤Hthresh, pixel (x, y) belongs to homogeneous area Domain；Work as Hd(i)>Hthresh, pixel (x, y) belongs to nonuniformity region；
(2). the laser image noise suppressed of fusion Multi-Scale Morphological Filtering result
The image denoising of Multiscale Morphological is to be weighted the shape filtering result of each mesostructure element averagely, If Y (x, y) is fused images, then：
Wherein, siFor the weight coefficient of each mesostructure element filtering image of correspondence；To siIt is defined as follows：
si'=| Hd(i)(x,y)-Hthresh|,
In (1) in step (two) of the present invention：
Judge image-region homogeneity threshold value HthreshSelection：
Calculate the area uniformity H of filtering image under out to out structural elementd(max)(x, y), using maximum between-cluster variance Method Otsu method carries out binaryzation to it, and the threshold value obtained in binarization is chosen for Hthresh。
The invention has the beneficial effects as follows：
(1) it is easy to hardware logic structure based on morphologic filtering method to realize, protects with energy real-time processing and preferably Hold the advantage of image border；
(2) morphologic filtering denoising is carried out using the structural element of multiple yardsticks, large-scale structure unit can be made full use of Plain filter effect is strong and small-scale structure element keeps edge details clearly advantage；
(3) using the homogeneity of image local area, as the fusion foundation of each mesostructure element filtering image, can be with Obtain more preferable laser image denoising effect.It is main to be tied using large scale morphologic filtering for the image-region of uniform gray level Really, obtaining stronger suppression noise immune；For the non-uniform image region comprising abundant details, mainly using little yardstick shape State filter result, to play little yardstick under locating features, obtain abundant edge.
Present invention could apply to there is speckle in Laser Active Imaging System Used, synthetic aperture radar, infrared medicine imaging etc. The image denoising occasion of noise.
Description of the drawings
Fig. 1 is the flow chart of the inventive method；
Fig. 2 is the laser image for shooting；
Fig. 3 is the area uniformity figure of out to out filtering image；
Fig. 4 is the noise suppressed result images of the inventive method；
Fig. 5 is Lee filtering images；
Fig. 6 is SBF filtering images；
Fig. 7 is Kuan filtering images.
Specific embodiment
Comprise the following steps：
(1), using different scale structural element, respectively form open-close is carried out to laser image and filtering is closed-opened to form Computing；
Morphologic basic operation has expansion, burn into open and close.Can derive and combine based on these basic operations Into various morphology practical algorithms.It is assumed that structural element is B, signal is F, and the basic operation of gray scale morphology definition is:
Expansion：
Corrosion：
(F Θ B) (x, y)=min F (x+s, y+t)-B (s, t) | (x+s), (y+t) ∈ DF；(s,t)∈DB} (2)
Open：
Closure：
Generally, the noise in image is often made up of the spike of lower convexity on signal.Noise is being carried out to image During suppression, the most frequently used Mathematical morphology filter wave technology is opening operation, closed operation and their hybrid operation.This paper algorithms lead to first Cross different order cascade open and close operator, structural configuration open-close and form close-drive wave filter；
" disk " structural element of different scale is chosen, radius is respectively d1、d2……dn, structural element Bd(i)Represent, Bd(1)For radius d1" disk " structural element, Bd(2)It is d for radius2" disk " structural element, by that analogy；If F (x, y) For input signal, S (x, y) is output signal, using structural element Bd(i)Make form open-close to input signal and form is closed-opened Computing, operation result is expressed as OCd(i)(x,y)、COd(i)(x,y)；
Form open-close and form are closed-opening operation result combines：
Wherein,Morphology opening operation is represented,Represent closing operation of mathematical morphology, OCCOd(i)(x, y) is represented and is utilized radius For di" disk " structural element carry out form open-close and form closes the-filter result image that combines of opening operation；
(2) according to the homogeneity of image local area, by the fusion of multi-scale morphology filtering result.
For the image-region of uniform gray level, large scale morphologic filtering result is mainly adopted, to obtain stronger suppression Noise immune；For the non-uniform image region comprising abundant details, mainly using little scale topographical filter result, to play Locating features under little yardstick, obtain abundant edge.
This process includes two steps：One judges that each pixel belongs to homogenous region or nonuniformity region in image； Two is the homogeneity according to image local area, and multi-scale morphology filtering result is merged, and detailed step is as follows：
(1). judge whether each pixel belongs to homogenous region in image
For each image pixel, surrounding M × M neighborhoods are chosen, according to formula (8), calculate uniformity H of the image-region (x,y)。
Wherein, f (x, y) represents the M × M neighborhoods centered on pixel (x, y), and Var (f (x, y)), E [f (x, y)] are respectively Represent the variance and average of grey scale pixel value in regional area；
For judging the threshold value of image-region homogeneity for Hthresh, work as Hd(i)≤Hthresh, pixel (x, y) belongs to homogeneous area Domain；Work as Hd(i)>Hthresh, pixel (x, y) belongs to nonuniformity region；
It is described below and judges image-region homogeneity threshold value HthreshSelection because the filtering figure of large-scale structure element As less by noise jamming, so calculating the H of filtering image under out to out structural elementd(max)(x, y), using Otsu methods (maximum variance between clusters) carry out binaryzation to it, and bianry image correspond to the marginal information of filtering image, binarization pair The threshold value answered just is to determine the foundation of image edge information, is also to determine whether pixel is located at the foundation of homogeneous area；So, choosing Select threshold value H of binarizationthreshWhether belong to the criterion of homogenous region as each pixel in image.
(2). the laser image noise suppressed of warm Multi-Scale Morphological Filtering result
The image denoising of Multiscale Morphological is to be weighted the shape filtering result of each mesostructure element averagely, If Y (x, y) is fused images, then
Wherein, siTo correspond to the weight coefficient of each mesostructure element filtering image, to siIt is defined as follows：
si'=| Hd(i)(x,y)-Hthresh|,
When pixel (x, y) belongs to homogenous region, i.e. Hd(i)≤Hthresh, the weight coefficient of large scale shape filtering image need to It is larger, can so make full use of the noiseproof feature of large-scale structure element, it is generally the case that for pixel (x, y), d I () yardstick is bigger, Hd(i)(x, y) value is less, according to formula (10), si' bigger, weight coefficient siAlso it is bigger；
When pixel (x, y) belongs to nonuniformity region, i.e. Hd(i)>Hthresh, the weight coefficient of little scale topographical filtering image need to It is larger, can so retain the marginal information in small-scale structure element filtering image.Under normal circumstances, for pixel (x, y), d (i) yardsticks are less, Hd(i)(x, y) value is bigger, according to formula (10), si' bigger, weight coefficient siAlso it is bigger.
The flow chart of the inventive method is as shown in Figure 1.
It is following with reference to specific experiment example further illustrating the present invention.
(1), the laser image of the building shot in experimental example of the present invention, as shown in Fig. 2 resolution ratio is 510*384, 256 grades of gray scales, choose " disk " structural element that radius is 2,3,4,5, using formula (5), (6), (7) Fig. 2 are carried out respectively many Scale topographical is filtered, and obtains four width filtering images；
(2), according to the homogeneity of image local area, by the morphologic filtering image co-registration of four width different scales, in detail Step is as follows：
1. it is 5*5 to choose neighborhood window, using formula (8), the area uniformity of four width filtering images, maximum chi is calculated respectively The area uniformity figure of degree filtering image, as shown in Figure 3.
2. the binary-state threshold H of Fig. 3 is obtained using Otsu methodsthreshIt is each in image in this, as judging for 0.015686 Whether pixel belongs to the standard of homogenous region.
3. the weight coefficient of four width different scale structural element filtering images of correspondence is calculated according to formula (10), according to formula (9) Four width filtering images are merged, final noise-reduced image is obtained, as shown in Figure 4.
Below several laser image denoising methods to commonly using have carried out emulation experiment, and Fig. 5, Fig. 6, Fig. 7 are respectively Lee filters Ripple algorithm, SBF filtering algorithms, the result of Kuan filtering algorithms.It may be seen that the inventive method solves laser figure As the obvious problem of noise.Compared to Lee filtering algorithms, SBF filtering algorithms, Kuan filtering algorithms, the inventive method can have Effect ground suppresses picture noise and keeps image definition.
Additionally, in order to quantitatively weigh filter effect, by calculating the speckle index of each image distinct methods are compared Suppression speckle noise ability.Speckle index is less, and the algorithm suppresses the ability of speckle noise stronger.Speckle Index Definition is：
Wherein, m, n represent image size, δijThe standard deviation of element in filter window is represented,Represent unit in filter window The average of element.The speckle index of original image and several algorithm process results is given in table 1.As seen from Table 1, the inventive method Speckle noise suppressing ability it is stronger.
The speckle index of the distinct methods result of table 1
Original image Lee is filtered SBF Kuan The inventive method
0.0757 0.0316 0.0505 0.0358 0.0300

Claims (2)

1. a kind of image noise suppression method of Laser active illuminated imaging, it is characterised in that comprise the following steps：
(1), using different scale structural element, carry out form open-close to the image of Laser active illuminated imaging respectively and form close- Open filtering operation；
" disk " structural element of different scale is chosen, radius is respectively d1、d2……dn, structural element Bd(i)Represent, Bd(1) It is d for radius1" disk " structural element, Bd(2)It is d for radius2" disk " structural element, by that analogy；If F (x, y) is Input signal, S (x, y) is output signal, using structural element Bd(i)Make form open-close to input signal and fortune is closed-opened to form Calculate, operation result is expressed as OCd(i)(x,y)、COd(i)(x,y)；
OCd(i)(x, y)=((F ο Bd(i))·Bd(i))(x,y) (5)
COd(i)(x, y)=((FBd(i))οBd(i))(x,y) (6)
Form open-close and form are closed-opening operation result combines：
OCCO d ( i ) ( x , y ) = 1 2 [ OC d ( i ) ( x , y ) + CO d ( i ) ( x , y ) ] - - - ( 7 )
Wherein, " ο " represents morphology opening operation, and " " represents closing operation of mathematical morphology, OCCOd(i)(x, y) is represented and is utilized radius as di " disk " structural element carry out form open-close and form closes the-filter result image that combines of opening operation；
(2) according to the homogeneity of image local area, by the fusion of multi-scale morphology filtering result；
(1). judge whether each pixel belongs to homogenous region in image
For each image pixel, surrounding M × M neighborhoods are chosen, according to formula (8), filtering image under each yardstick is calculated respectively OCCOd(i)The area uniformity of (x, y), is as a result designated as Hd(i)(x,y)；
H ( x , y ) = V a r ( f ( x , y ) ) E [ f ( x , y ) ] - - - ( 8 )
Wherein, f (x, y) represents the M × M neighborhoods centered on pixel (x, y), and Var (f (x, y)), E [f (x, y)] are represented respectively The variance and average of grey scale pixel value in regional area；
For judging the threshold value of image-region homogeneity for Hthresh, work as Hd(i)≤Hthresh, pixel (x, y) belongs to homogenous region；When Hd(i)>Hthresh, pixel (x, y) belongs to nonuniformity region；
(2). the laser image noise suppressed of fusion Multi-Scale Morphological Filtering result
The image denoising of Multiscale Morphological is to be weighted the shape filtering result of each mesostructure element averagely, if Y (x, y) is fused images, then：
Y ( x , y ) = Σ i = 1 n s i · OCCO d ( i ) ( x , y ) - - - ( 9 )
Wherein, siFor the weight coefficient of each mesostructure element filtering image of correspondence；To siIt is defined as follows：
s i ′ = | H d ( i ) ( x , y ) - H t h r e s h | , s i = s i ′ Σ i = 1 n s i ′ - - - ( 10 ) .
2. a kind of image noise suppression method of Laser active illuminated imaging according to claim 1, it is characterised in that step (2) in (1) in：
Judge image-region homogeneity threshold value HthreshSelection：
Calculate the area uniformity H of filtering image under out to out structural elementd(max)(x, y), using maximum variance between clusters Otsu methods carry out binaryzation to it, and the threshold value obtained in binarization is chosen for Hthresh。
CN201610850521.5A 2016-09-26 2016-09-26 A kind of image noise suppression method of Laser active illuminated imaging Active CN106651781B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610850521.5A CN106651781B (en) 2016-09-26 2016-09-26 A kind of image noise suppression method of Laser active illuminated imaging

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610850521.5A CN106651781B (en) 2016-09-26 2016-09-26 A kind of image noise suppression method of Laser active illuminated imaging

Publications (2)

Publication Number Publication Date
CN106651781A true CN106651781A (en) 2017-05-10
CN106651781B CN106651781B (en) 2019-05-24

Family

ID=58853936

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610850521.5A Active CN106651781B (en) 2016-09-26 2016-09-26 A kind of image noise suppression method of Laser active illuminated imaging

Country Status (1)

Country Link
CN (1) CN106651781B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109712129A (en) * 2018-12-25 2019-05-03 河北工业大学 A kind of arc image processing method based on mathematical morphology
CN112330553A (en) * 2020-10-30 2021-02-05 武汉理工大学 Crack image denoising method, device and storage medium
CN113298764A (en) * 2021-05-11 2021-08-24 合肥富煌君达高科信息技术有限公司 High-speed camera imaging quality analysis method based on image noise analysis

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102736085A (en) * 2012-06-21 2012-10-17 中国科学院半导体研究所 Image target seeking laser imaging distance measurement method and device

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102736085A (en) * 2012-06-21 2012-10-17 中国科学院半导体研究所 Image target seeking laser imaging distance measurement method and device

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
李晓峰等: "激光主动成像图像噪声分析与抑制", 《红外与激光工程》 *
王宇等: "一种提高计算机重构三维图像质量的方法", 《计算机仿真》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109712129A (en) * 2018-12-25 2019-05-03 河北工业大学 A kind of arc image processing method based on mathematical morphology
CN112330553A (en) * 2020-10-30 2021-02-05 武汉理工大学 Crack image denoising method, device and storage medium
CN112330553B (en) * 2020-10-30 2022-07-01 武汉理工大学 Crack image denoising method, device and storage medium
CN113298764A (en) * 2021-05-11 2021-08-24 合肥富煌君达高科信息技术有限公司 High-speed camera imaging quality analysis method based on image noise analysis

Also Published As

Publication number Publication date
CN106651781B (en) 2019-05-24

Similar Documents

Publication Publication Date Title
CN108053417B (en) lung segmentation device of 3D U-Net network based on mixed rough segmentation characteristics
CN101661611B (en) Realization method based on bayesian non-local mean filter
CN106709928A (en) Fast noise-containing image two-dimensional maximum between-class variance threshold value method
CN103473767B (en) The method and system that a kind of soft tissues of abdomen nuclear-magnetism image is cut apart
CN103208115B (en) Based on the saliency method for detecting area of geodesic line distance
CN104574445A (en) Target tracking method and device
CN101996406A (en) No-reference structure definition image quality evaluation method
CN101685158B (en) Hidden Markov tree model based method for de-noising SAR image
CN104103080A (en) Method of small dim target detection under complicated background
CN101901476A (en) SAR image de-noising method based on NSCT domain edge detection and Bishrink model
Zhang et al. Decision-based non-local means filter for removing impulse noise from digital images
CN109635789B (en) High-resolution SAR image classification method based on intensity ratio and spatial structure feature extraction
Nadeem et al. Fuzzy logic based computational model for speckle noise removal in ultrasound images
CN102096913B (en) Multi-strategy image fusion method under compressed sensing framework
CN102073994B (en) Ultrasonic medical image speckle noise inhibition method based on multi-scale anisotropic diffusion
CN103400368A (en) Parallel rapid SAR image segmentation method based on graph theory and superpixel
CN107194896A (en) A kind of background suppression method and system based on neighbour structure
CN106651781A (en) Image noise suppression method for laser active imaging
CN104021536A (en) Self-adaptation SAR image and multispectral image fusion method
CN104680536A (en) Method for detecting SAR image change by utilizing improved non-local average algorithm
CN108830883B (en) Visual attention SAR image target detection method based on super-pixel structure
Zhou et al. New improved optimized method for medical image enhancement based on modified shark smell optimization algorithm
Yan et al. Improved mask R-CNN for lung nodule segmentation
CN105469428A (en) Morphological filtering and SVD (singular value decomposition)-based weak target detection method
CN103093432A (en) Polarized synthetic aperture radar (SAR) image speckle reduction method based on polarization decomposition and image block similarity

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