CN108038856B - Infrared small target detection method based on improved multi-scale fractal enhancement - Google Patents

Infrared small target detection method based on improved multi-scale fractal enhancement Download PDF

Info

Publication number
CN108038856B
CN108038856B CN201711403183.1A CN201711403183A CN108038856B CN 108038856 B CN108038856 B CN 108038856B CN 201711403183 A CN201711403183 A CN 201711403183A CN 108038856 B CN108038856 B CN 108038856B
Authority
CN
China
Prior art keywords
scale
formula
target
image
value
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
Application number
CN201711403183.1A
Other languages
Chinese (zh)
Other versions
CN108038856A (en
Inventor
谷雨
彭冬亮
冯秋晨
刘俊
陈华杰
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.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
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 Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN201711403183.1A priority Critical patent/CN108038856B/en
Publication of CN108038856A publication Critical patent/CN108038856A/en
Application granted granted Critical
Publication of CN108038856B publication Critical patent/CN108038856B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • 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
    • 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

Abstract

The invention relates to an infrared small target detection method based on improved multi-scale fractal enhancement. The traditional fractal feature enhancement-based infrared small target detection algorithm has the problem of high calculation complexity. The method utilizes the difference between the maximum value and the minimum value of the pixels in the multi-scale region as a measure, then adopts the region average value to obtain the multi-scale fractal feature vector of each point of the pixels, evaluates the significance of each point of the pixels through a defined new significance measure criterion, and finally carries out target detection through an adaptive threshold segmentation algorithm based on the enhanced image. The method provided by the invention reduces the calculation amount of the algorithm and improves the detection speed.

Description

Infrared small target detection method based on improved multi-scale fractal enhancement
Technical Field
The invention belongs to the field of infrared small target detection, and relates to an infrared small target detection method based on improved multi-scale fractal enhancement.
Background
The infrared weak and small target detection is one of key technologies of an infrared searching and tracking system, can effectively improve the monitoring range, and plays an important role in the fields of navigation, air defense, safety monitoring and the like. Two difficulties exist in the detection of infrared small and weak targets: (1) small objects have no obvious texture and shape characteristics; (2) under the influence of background target radiation and an image sensor, random noise and a large number of clutter exist in an infrared image, and the signal-to-noise ratio of the image is low. Despite the numerous algorithms proposed, the efficient detection of small targets in complex environments remains an issue that has not yet been fully resolved.
Infrared small target detection techniques can be broadly divided into two categories: pre-track Detection (DBT) and pre-detection Tracking (TBD). The TBD technology uses multi-frame images for accumulation to detect a weak target, and the DBT is to position the target in the first frame image where the target appears by using a target detection algorithm and then estimate the position of the target by using a tracking technology by using the space-time consistency of the target in continuous images. The small target is detected in a single image without mastering the prior knowledge of the target, and the small target detection method is high in calculation efficiency and is commonly used for some detection equipment. Although much research has focused on DBT technology over the last two decades, effective infrared small target detection algorithms have been proposed, such as TopHat Filtering, MaxMean, and MaxMedian, among others.
At present, most of algorithms proposed by researchers are based on the assumption that the gray scale contrast of a target and the surrounding environment thereof is greater than that of a background region, and firstly, an input image is enhanced, and then, a threshold value is used for segmentation to obtain a candidate region of the target to be detected. According to the fractal geometric theory, because a natural target and an artificial target have different internal structures, a fractal model is more suitable for the natural target of mountain, cloud, water, plant and the like in a certain specific range, but is not suitable for the artificial target, so that effective multi-scale fractal characteristics can be designed based on the fractal theory, an image is enhanced first, and then the detection of the infrared small target is realized based on a self-adaptive threshold segmentation method. For example, an average gray scale difference absolute value maximum mapping algorithm (AGADMM) defines target regions with different scales with a pixel coordinate of each point as a central point, and also defines a background region with a larger scale, and evaluates the significance of each target region and the background region by calculating the maximum value of the difference absolute value of the gray scale mean value of the pixels, and uses the maximum value as a measure to enhance the image, and an adaptive threshold segmentation algorithm is used to realize target detection based on the enhanced image.
Disclosure of Invention
The method provided by the invention considers the problem of high computational complexity of the traditional fractal feature enhancement-based infrared small target detection algorithm, provides a new multi-scale fractal feature-based significance measurement criterion by analyzing the existing algorithm, simplifies the algorithm, designs an improved multi-scale fractal enhancement-based infrared small target detection method, reduces the calculated amount of the algorithm, and improves the real-time property on the premise of ensuring the target detection rate. The method utilizes the difference between the maximum value and the minimum value of the pixels in the multi-scale region as a measure, then adopts the region average value to obtain the multi-scale fractal feature vector of each point of the pixels, evaluates the significance of each point of the pixels through a defined new significance measure criterion, and finally carries out target detection through an adaptive threshold segmentation algorithm based on the enhanced image. The designed target detection algorithm can obtain the enhanced image by only applying simple four-rule operation, is simple to realize, is very suitable for embedded application, and improves the real-time property of the algorithm on the premise of ensuring the detection rate.
The technical scheme adopted by the invention comprises the following steps:
and (1) acquiring a multi-scale fractal feature vector of each pixel in an original infrared image I.
And (2) calculating the significance of each pixel through a defined new significance measurement criterion based on the multi-scale fractal feature vector to obtain an enhanced image.
And (3) based on the enhanced image, performing target segmentation by adopting a self-adaptive threshold segmentation algorithm to obtain a detected target.
Compared with the prior art, the invention has the following remarkable advantages: (1) when the image enhancement is carried out, the parameters needing to be set are few, and only the maximum scale of the multi-scale fractal features needs to be set and calculated. With the increase of the maximum scale, the broadening effect of the target in the enhanced image is more obvious, so that the visual and threshold segmentation of small targets are facilitated; (2) the contrast information of the target local area is effectively utilized, most targets can be detected under the condition of less false alarms, and the designed method can simultaneously detect the bright targets and the dark targets in the image. (3) The method is not only suitable for small target detection, but also suitable for large target detection including SAR images with strong speckle noise through designing reasonable scale and detection threshold. (4) The traditional algorithm is complex in calculation, the algorithm only relates to four arithmetic operations, the implementation in an embedded system is convenient, the calculation amount of the algorithm is low, and the real-time performance is good.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a graph of the effect of maximum scale on infrared image enhancement;
FIG. 3 infrared image enhancement and detection results (small targets) based on improved multi-scale fractal enhancement; wherein (a) - (f) are infrared images under different scenes;
FIG. 4 infrared image enhancement and detection results (large targets) based on improved multi-scale fractal enhancement; wherein (a) - (f) are infrared images under different scenes;
FIG. 5 SAR image enhancement and detection results (large target) based on improved multi-scale fractal enhancement; wherein (a) - (d) are infrared images under different scenes;
FIG. 6 compares the enhancement effect of five other exemplary infrared small target detection algorithms; wherein (1) to (6) are infrared images under different scenes; (a) - (d) target enhancement effect for original image and different detection algorithm.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
As shown in FIG. 1, the method comprises the following steps:
and (1) acquiring a multi-scale fractal feature vector of each pixel in an original infrared image I. The method comprises the following specific steps:
according to the fractal geometry theory, the relationship between fractal measure and scale can be described as follows by Richardson's law:
M()＝Kd-FDformula (1)
Where, denotes a scale, 1, 2., M () denotes a measure at the scale, FD and d denote a fractal dimension and a topological dimension, respectively, and K is a fractal parameter. For a two-dimensional grayscale image can be described as:
A(x,y,)＝K(x,y,)2-FD(x,y,)formula (2)
Where a (x, y,) represents a surface area measure of the image grey scale surface at scale.
When the measurement scales are respectively1,2Then, it can be obtained from the formula (2),
logA(x,y,1)＝(2-FD(x,y,1))log(1)+logK(x,y,1) Formula (3)
logA(x,y,2)＝(2-FD(x,y,2))log(2)+logK(x,y,2) Formula (4)
According to the theory of fractal geometry, it can be known that for an ideal fractal, the fractal dimension FD is a quantity independent of all scales and is always a constant. Therefore, it is assumed that FD in equations (3) and (4) is constant at different scales of measurement, and is set1＝,2When the measurement scale is-1, the corresponding D-dimensional area K (x, y,) can be expressed as
Figure GDA0001602940250000031
When A (x, y,) is calculated by carpet overlay, there are
Figure GDA0001602940250000041
Wherein, V (x, y,) is the volume at coordinate (x, y) under the scale, the calculation formula is shown as formula (7), and U (x, y,) and B (x, y,) are the maximum value and the minimum value of the pixel in the neighborhood range at coordinate (x, y) under the scale, respectively.
Figure GDA0001602940250000042
Substituting formula (6) for formula (5) to obtain
Figure GDA0001602940250000043
By analyzing the formula (7) and the formula (8), the volume V of each pixel of the homogeneous background area is small and is close to 0 under different scales; in the inhomogeneous background region, although the value of V is large, the change rate is relatively large under different scales due to the fluctuation of the background, i.e. the absolute value of the second part in the formula (8) is large, so the value of K is reduced; in the target region, the maximum value or the minimum value of the local region is fixed, Vmax is small, and the change rate is small, so that the K value is maximum in the target region.
Based on the above analysis, if the last two parts in equation (8) are omitted, K (x, y') ═ V (x, y, -1) can be used, and the sum of the differences between the maximum value and the minimum value of the pixels in one region can be used as a measure of the saliency of the region centered on the coordinate point (x, y).
Based on the above analysis, let the scale be 2 …maxAnd when 1, the original infrared image,maxis the largest dimension. In the original infrared image I, the pixel coordinate of a point in the image I is (x, y), the difference value between the maximum value and the minimum value of the pixel value in the neighborhood range of the coordinate point under the scale is calculated through a formula (9) and is used as the contrast of the pixel of the point under the scale, and a contrast image I is obtained; in order to eliminate the influence of the cumulative effect brought by different scales, calculating the average contrast of each point pixel under the scale through a formula (10) to obtain a fractal feature vector under the scale; and further taking the scale as a variable to obtain a multi-scale fractal feature vector t (x, y, the following) which is a fractal feature vector.
Figure GDA0001602940250000044
Figure GDA0001602940250000051
And (2) calculating the significance of each pixel based on the multi-scale fractal feature vector t (x, y, in the following) to obtain an enhanced image. The method comprises the following specific steps:
based on the analysis of equation (8), for the target region, the mean value of the multi-scale fractal feature vector defined by equation (10) is large and the variation is small, while for the background region, especially for the non-uniform background region, although the mean value is also large, the variation is large, so the significance metric value at the coordinate point (x, y) is calculated by using equation (11), and the enhanced image is obtained and recorded as E (x, y).
E(x,y)＝mean(t(x,y,:))2-std(t(x,y,:))2Formula (11)
Where mean () and std () are functions of the mean and standard deviation, respectively.
And (3) based on the enhanced image, performing target segmentation by adopting a self-adaptive threshold segmentation algorithm to obtain a detected target. The method comprises the following specific steps:
using adaptive thresholdsThe segmentation algorithm is used for target detection, and in order to eliminate the influence of boundary effect, 2 times of the maximum scale is removed when the segmentation threshold is calculatedmaxThe specific calculation formula of the boundary region is as follows:
μ＝mean(E(2×max+1:rows-2×max,2×max+1:cols-2×max) Formula (12)
＝std(E(2×max+1:rows-2×max,2×max+1:cols-2×max) Formula (13)
PSR ═ 255- μ)/formula (14)
T ═ c × PSR × + μ formula (15)
D (x, y) ═ E (x, y) ≥ T formula (16)
Wherein mu is the pixel mean value of the enhanced image after the edge area is removed, and is the standard deviation of the pixel distribution of the enhanced image after the edge area is removed; PSR is peak sidelobe ratio; c is a segmentation coefficient, and the value range can be set to [ 0.50.65 ] for small target detection and [ 0.150.45 ] for large target detection. rows and cols are the height and width of the original infrared image I; t is a threshold for object detection and D is the detection of a pixel belonging to an object.
To verify the effectiveness of the present invention, the maximum scale was first analyzed experimentallymaxThe influence on the enhancement effect of the input infrared image can be seen from fig. 2, as the scale is increased, the target in the enhanced image has an obvious broadening effect, and the greater the maximum scale is, the more obvious the broadening effect is, the contrast of the target in the enhanced infrared image is improved, and the visual effect is enhanced obviously. The performance of the method is tested by 6 groups of infrared images with small targets, and the enhanced images and the detection result are shown in figure 3. As can be seen from fig. 3, in addition to the detection of a bright target, a suspicious dark target can be detected as well. When a large target exists in the scene, by selecting an appropriate division coefficient c, and setting c to 0.35, the target detection results for the infrared image and the SAR image are shown in fig. 4 and 5. Even if the SAR image has strong speckle noise, the proposed algorithm effectively utilizes the contrast information of the target local area, and can detect most of the SAR images under the condition of less false alarmsAnd (4) a target.
FIG. 6 is a comparison of the target enhancement effect of the proposed method of the present invention with five typical infrared small target detection algorithms such as TopHat filtration, MaxMean, MaxMedian, AGADMM and NWIE. Wherein the NWIE algorithm is an AGADMM algorithm based on local entropy weighting. As can be seen from fig. 6, the method provided by the present invention has an obvious effect of enhancing a small target, and meanwhile, for a larger target situation existing in an image, such as a bridge in which a target in the fifth row in fig. 6 is a submarine which just floats out of the water surface, the target area is divided into a plurality of small point targets after being enhanced by other methods, and the method can completely enhance the bridge part of the submarine, which also indicates that the method is not only suitable for small target detection, but also suitable for image detection including a large target.
In order to verify the real-time performance of the invention, the following PC hardware configuration is adopted: the CPU is Intel (R) core (TM) i5-3230M @2.6GHz, the memory is 12GB, the video card is NVIDIA NVS5400M, and the 2G independent video memory; the algorithm is implemented by Matlab and C + +. The target detection experiment was performed at a maximum scale of 4, and the time required for detection of infrared images at different resolutions is shown in table 1. Experiments prove that the method can meet the real-time requirement and is very suitable for the implementation of an embedded system.
TABLE 1 time (ms) required for detection at different image resolutions according to the invention
Serial number Image type Resolution (Pixel) Invention algorithm Multi-scale fractal feature algorithm
1 IR 200×150 14.42 20.70
2 IR 280×228 28.40 39.37
3 IR 250×200 21.71 31.93
4 IR 281×240 29.41 40.49
5 IR 220×140 14.60 18.55
6 IR 320×240 33.51 48.16

Claims (4)

1. The infrared small target detection method based on improved multi-scale fractal enhancement is characterized by comprising the following specific steps:
the method comprises the following steps of (1) obtaining a multi-scale fractal feature vector of each pixel in an original infrared image I';
step (2), calculating the significance of each pixel based on the multi-scale fractal feature vector, and obtaining an enhanced image E (x, y) according to a formula (1);
E(x,y)＝mean(t(x,y,:))2-std(t(x,y,:))2formula (1)
Wherein mean () and std () are respectively the functions of mean value calculation and standard deviation, and (x, y) are pixel coordinates in the original infrared image I'; t (x, y:) represents a multi-scale fractal feature vector of pixel coordinates (x, y);
step (3) based on the enhanced image, adopting a self-adaptive threshold segmentation algorithm to segment the target, and obtaining the detected target according to formulas (2) to (6);
μ＝mean(E(2×max+1:rows-2×max,2×max+1:cols-2×max) Formula (2)
＝std(E(2×max+1:rows-2×max,2×max+1:cols-2×max) Formula (3)
PSR ═ 255- μ)/formula (4)
T ═ c × PSR × + μ formula (5)
D (x, y) ═ E (x, y) ≥ T formula (6)
Wherein mu is the pixel mean value of the enhanced image after the edge area is removed, and is the standard deviation of the pixel distribution of the enhanced image after the edge area is removed; PSR is peak sidelobe ratio; c represents a small target segmentation coefficient; rows and cols are respectively the height and width of the original infrared image I'; t is a threshold value for target detection; d is the detected pixel belonging to the target;
the step (1) is specifically as follows:
according to the fractal geometric theory, deducing: when the measurement scale is "K", the corresponding D-dimensional area K (x, y,) can be expressed as
Figure FDA0002459343330000011
Wherein a (x, y,) represents a surface area measure of the image gray scale surface at scale;
when A (x, y,) is calculated by carpet overlay, there are
Figure FDA0002459343330000012
V (x, y,) is the volume at the coordinate (x, y) under the scale, the calculation formula is shown as formula (9), U (x, y,) and B (x, y,) are the maximum value and the minimum value of the pixels in the neighborhood range at the coordinate (x, y) under the scale respectively;
Figure FDA0002459343330000021
substituting formula (8) for formula (7) to obtain
Figure FDA0002459343330000022
According to the analysis of the formula (9) and the formula (10), under different scales, the volume V of each pixel of the homogeneous background area is small and is close to 0; in the inhomogeneous background region, although the value of V is large, the change rate is relatively large under different scales due to the fluctuation of the background, i.e. the absolute value of the second part in the formula (10) is large, so the value of K is reduced; in the target area, no matter whether the target is a bright target or a dark target, the maximum value or the minimum value of the local area is fixed, Vmax is obtained, the change rate is small, and therefore the K value is maximum in the target area;
based on the above analysis, if the last two parts in equation (10) are omitted, K (x, y,) V (x, y, -1) is present, and the sum of the differences between the maximum value and the minimum value of the pixels in one region can be used as the saliency metric of the region centered on the coordinate point (x, y);
based on the above analysis, let the scale be 2 …maxAnd when 1, the original infrared image,maxis the maximum dimension; in the original infrared image IThe pixel coordinate of one point is (x, y), the difference value between the maximum value and the minimum value of the pixel value in the neighborhood range of the coordinate point under the scale is calculated by the formula (11) and is used as the contrast of the pixel of the point under the scale, and a contrast image I is obtained; in order to eliminate the influence of the cumulative effect brought by different scales, calculating the average contrast of each point pixel under the scale through a formula (12) to obtain a fractal feature vector under the scale; further, the scale is taken as a variable to obtain a multi-scale fractal feature vector t (x, y,: r);
Figure FDA0002459343330000023
Figure FDA0002459343330000031
2. the method according to claim 1, characterized in that the partition coefficient c in the formula (5) is adjusted to a large target partition coefficient.
3. The method of claim 1 or 2, characterized in that the original image is replaced by an infrared image into an SAR image, and the segmentation coefficient c in formula (5) is adjusted to the SAR image small target segmentation coefficient.
4. The method of claim 1 or 2, characterized in that the original image is replaced by the SAR image from the infrared image, and the segmentation coefficient c in formula (5) is adjusted to the SAR image large target segmentation coefficient.
CN201711403183.1A 2017-12-22 2017-12-22 Infrared small target detection method based on improved multi-scale fractal enhancement Active CN108038856B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711403183.1A CN108038856B (en) 2017-12-22 2017-12-22 Infrared small target detection method based on improved multi-scale fractal enhancement

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711403183.1A CN108038856B (en) 2017-12-22 2017-12-22 Infrared small target detection method based on improved multi-scale fractal enhancement

Publications (2)

Publication Number Publication Date
CN108038856A CN108038856A (en) 2018-05-15
CN108038856B true CN108038856B (en) 2020-08-04

Family

ID=62100448

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711403183.1A Active CN108038856B (en) 2017-12-22 2017-12-22 Infrared small target detection method based on improved multi-scale fractal enhancement

Country Status (1)

Country Link
CN (1) CN108038856B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109461164A (en) * 2018-09-21 2019-03-12 武汉大学 A kind of infrared small target detection method based on direction nuclear reconstitution
CN110084778B (en) * 2019-01-31 2021-04-13 电子科技大学 Infrared imaging cirrus cloud detection method based on fractal dictionary learning
CN110310264B (en) * 2019-06-25 2021-07-20 北京邮电大学 DCNN-based large-scale target detection method and device
CN113343758B (en) * 2021-04-26 2022-03-15 西安卓越视讯科技有限公司 Long-distance unmanned aerial vehicle small target detection method based on infrared image

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5671294A (en) * 1994-09-15 1997-09-23 The United States Of America As Represented By The Secretary Of The Navy System and method for incorporating segmentation boundaries into the calculation of fractal dimension features for texture discrimination
EP1708140A2 (en) * 2005-03-31 2006-10-04 Lockheed Martin Corporation Unresolved target detection improvement by use of multiple matched filters approach at different spatial phases
CN102521831A (en) * 2011-12-02 2012-06-27 南京信息工程大学 Robot vision image segmentation method based on multi-scale fractal dimension and neural network
CN103077534A (en) * 2012-12-31 2013-05-01 南京华图信息技术有限公司 Space-time multi-scale moving target detection method
CN103218782A (en) * 2013-04-11 2013-07-24 杭州电子科技大学 Infrared image strengthening method based on multiscale fractal characteristics
CN105741253A (en) * 2016-01-27 2016-07-06 北京理工大学 Enhancement estimation method of image fractal feature on the basis of merge replication

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8615133B2 (en) * 2007-03-26 2013-12-24 Board Of Regents Of The Nevada System Of Higher Education, On Behalf Of The Desert Research Institute Process for enhancing images based on user input
CN104501959B (en) * 2014-12-30 2016-08-17 华中科技大学 A kind of infared spectrum association intelligent detection method and device

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5671294A (en) * 1994-09-15 1997-09-23 The United States Of America As Represented By The Secretary Of The Navy System and method for incorporating segmentation boundaries into the calculation of fractal dimension features for texture discrimination
EP1708140A2 (en) * 2005-03-31 2006-10-04 Lockheed Martin Corporation Unresolved target detection improvement by use of multiple matched filters approach at different spatial phases
CN102521831A (en) * 2011-12-02 2012-06-27 南京信息工程大学 Robot vision image segmentation method based on multi-scale fractal dimension and neural network
CN103077534A (en) * 2012-12-31 2013-05-01 南京华图信息技术有限公司 Space-time multi-scale moving target detection method
CN103218782A (en) * 2013-04-11 2013-07-24 杭州电子科技大学 Infrared image strengthening method based on multiscale fractal characteristics
CN105741253A (en) * 2016-01-27 2016-07-06 北京理工大学 Enhancement estimation method of image fractal feature on the basis of merge replication

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于红外图像的内河运动船舶目标检测和跟踪技术研究;刘俊;《中国博士学位论文全文数据库（信息科技辑）》;20090615(第06期);第I138-45页 *

Also Published As

Publication number Publication date
CN108038856A (en) 2018-05-15

Similar Documents

Publication Publication Date Title
CN108038856B (en) Infrared small target detection method based on improved multi-scale fractal enhancement
CN106856002B (en) Unmanned aerial vehicle shooting image quality evaluation method
YongHua et al. Study on the identification of the wood surface defects based on texture features
CN109740445B (en) Method for detecting infrared dim target with variable size
CN109886218B (en) SAR image ship target detection method based on superpixel statistic dissimilarity
Li et al. Infrared maritime dim small target detection based on spatiotemporal cues and directional morphological filtering
CN109086724B (en) Accelerated human face detection method and storage medium
CN105335965B (en) Multi-scale self-adaptive decision fusion segmentation method for high-resolution remote sensing image
Qi et al. Small infrared target detection utilizing local region similarity difference map
Fu et al. Infrared sea-sky line detection utilizing self-adaptive Laplacian of Gaussian filter and visual-saliency-based probabilistic Hough transform
CN113205494B (en) Infrared small target detection method and system based on adaptive scale image block weighting difference measurement
CN107369163B (en) Rapid SAR image target detection method based on optimal entropy dual-threshold segmentation
CN112163606B (en) Infrared small target detection method based on block contrast weighting
Zhengzhou et al. Gray-scale edge detection and image segmentation algorithm based on mean shift
CN108830864A (en) Image partition method
CN110321808B (en) Method, apparatus and storage medium for detecting carry-over and stolen object
CN115049552A (en) Infrared small target detection method based on structure tensor weighted local contrast measurement
CN108401563B (en) Infrared small target detection method based on Multiscale mean values filtering and conspicuousness detection
Yuanyuan et al. Infrared small dim target detection using local contrast measure weighted by reversed local diversity
Tian et al. Joint spatio-temporal features and sea background prior for infrared dim and small target detection
CN112766032A (en) SAR image saliency map generation method based on multi-scale and super-pixel segmentation
CN112348853A (en) Particle filter tracking method based on infrared saliency feature fusion
Li et al. An Infrared small target detection method based on local contrast measure and gradient property
Wei et al. Small infrared target detection based on image patch ordering
CN110674782A (en) Infrared dim target detection method based on fractional entropy

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