CN104200450B - A kind of thermal-induced imagery definition enhancing method - Google Patents

A kind of thermal-induced imagery definition enhancing method Download PDF

Info

Publication number
CN104200450B
CN104200450B CN201410421814.2A CN201410421814A CN104200450B CN 104200450 B CN104200450 B CN 104200450B CN 201410421814 A CN201410421814 A CN 201410421814A CN 104200450 B CN104200450 B CN 104200450B
Authority
CN
China
Prior art keywords
image
focal length
short focus
mrow
fused
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.)
Expired - Fee Related
Application number
CN201410421814.2A
Other languages
Chinese (zh)
Other versions
CN104200450A (en
Inventor
杜娟
冯颖
胡池
胡跃明
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN201410421814.2A priority Critical patent/CN104200450B/en
Publication of CN104200450A publication Critical patent/CN104200450A/en
Application granted granted Critical
Publication of CN104200450B publication Critical patent/CN104200450B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Image Processing (AREA)

Abstract

The invention discloses a kind of thermal-induced imagery definition enhancing method, comprise the following steps:A gathers thermal-induced imagery, and carries out enhancing contrast pretreatment;Infrared chart seems by two parallel optical axis structures and the thermal infrared imager with different focal is obtained, and thermal-induced imagery includes short focus image and focal length image;B expands short focus image, and focal length image is divided into low frequency and high frequency imaging;C is using DLT algorithms and translation mapping table, the mapping relations that the short focus image set up after expanding aligns with the plane at infinity of original focal length image;D sets up image pyramid, calculates the horizontal direction parallax between image;The HFS of focal length image is fused in the short focus image of expansion and is partly fused in the short focus image of expansion by E.The present invention reduces cost compared to high-resolution thermal infrared imager;The definition of regional area is added while ensure that large viewing scope.

Description

A kind of thermal-induced imagery definition enhancing method
Technical field
The present invention relates to infrared image processing field, and in particular to a kind of thermal-induced imagery definition enhancing method.
Background technology
Have tonal range low for thermal-induced imagery, general image is partially dark, blurred edges, the spy of a variety of noises of aliasing Point, the pretreatment to thermal-induced imagery is extremely important, and main task includes reduction noise, lifts tonal range, enhancing edge etc. Task.Due to the characteristic sensor of thermal infrared imager so that its imaging resolution is relatively low, the resolution ratio of general commercial thermal imaging system is all Not over 640*480.And the thermal imaging system of higher resolution is expensive, widespread adoption is difficult at this stage.Therefore it can examine Worry is merged using the image of low resolution, and more preferable definition is obtained with less expensive price.
Usual one secondary infrared image is made up of foreground and background, and the image-region often seen needed for observer is simultaneously It is not entire image, but, it is necessary to obtain local clear display based on main body region.
The content of the invention
In order to overcome the shortcoming and deficiency that prior art is present, the present invention provides a kind of infrared chart image sharpness enhancing side Method.
The present invention is adopted the following technical scheme that:
A kind of thermal-induced imagery definition enhancing method, comprises the following steps:
A gathers thermal-induced imagery, and carries out enhancing contrast pretreatment;The infrared chart seems by two directional lights Axle construction and the thermal infrared imager acquisition with different focal, the thermal-induced imagery include short focus image and focal length image;
B expands short focus image, and focal length image is divided into low frequency and high frequency imaging;
C is put down using DLT algorithms and translation mapping table, the infinity of the short focus image set up after expanding and original focal length image In face of neat mapping relations;
D sets up image pyramid, calculates the horizontal direction parallax between image;
The HFS of focal length image is fused in the short focus image of expansion by E.
The A specifically includes following steps:
A1 carries out smoothing denoising processing using median filtering algorithm or bilateral filtering algorithm to image;
A2 strengthens image high frequency detail using Butterworth sharpening filter, lifts image overall intensity.
The B specifically includes following steps:
B1 expands pretreated short focus image so that after expansion according to telephoto lens and the focal length ratio of short-focus lens Short focus image and focal length image there is identical ratio scale in public domain;
B2 uses gauss low frequency filter, and pretreated focal length image is divided into two images of high and low frequency, filtering Target be to make the fog-level of low-frequency image suitable with the short focus image that A1 is obtained.
The C specifically includes following steps:
Two camera lenses of C1 are according to parallel optical axis structure lay out in parallel, and the direction 1-2m sampled in Vertical camera lenses places one piece and put down Multiple spot lights are uniformly placed on panel, plate;
The short focus image that C2 is expanded B1, it is uniform on image with original focal length image as one group of demarcation photo Find out and assemble point coordinates, preferably at least more than 10 groups, and record more;
C3 utilizes above-mentioned multigroup match point, and description short focus image and original focal length are calculated using straight linear transfer algorithm Image is based on outerplanar homography matrix described in surface plate;
C4 calculates short focus image and original on the basis of the short focus image expanded using homography matrix and bilinear interpolation algorithm The coordinate map of beginning focal length image;
C5 removes surface plate described in C1, and multiple spot lights are then placed in the plane more than 10m, by calculating parallax, Obtain the coordinate map of plane at infinity alignment.
The D specifically includes following steps:
D1 is according to obtained coordinate map, in the short focus image of expansion, determines a region to be fused, obtains to be fused The coordinate starting point in region and the size in region to be fused, i.e., obtain the public domain between low-frequency image with B2;
D2 as image I, using low-frequency image as image J, on the basis of coordinate map, calculates in region to be fused Go out the disparity vector of each points to be matched of image I, form a disparity vector table.
The E specifically includes following steps:
Coordinate map and disparity vector table are done add operation and obtain comprehensive mapping table by E1, the short focus expanded as description The matching relationship of image and original focal length image;
The coordinate starting point in E2 regions to be fused starts, using comprehensive mapping table and bilinear interpolation algorithm, high frequency imaging It is fused in region to be fused, increases the definition of public domain.
The D2 specific steps include:
Three layers of gold as image I, using low-frequency image as image J, are set up in region to be fused by D21 to above-mentioned image respectively The image sequence of word tower structure, described three layers golden word layer include upper, middle and lower layer, the upper strata pyramid length and a width of lower floor gold Word tower length and wide 1/2;
The pixel on D22 traversal I images upper strata, sets up 3*3 characteristic block, according to coordinate map, on image J Horizontal unilateral direction is scanned for, and limits its hunting zone according to the experience of depth and parallax relation, poor using square-error That is SSD as weigh similarity indices,
(i', j ') is that coordinate points (i, j) draw its corresponding point coordinates on image J by inquiring about mapping table, so that really Fixed a number of thick match point parallactic shift amount;
D23, using SSD as similarity evaluation index, is directly transferred to lower floor and carries out surrounding according to the thick match point of determination The accurate of four points is compared, until it is determined that parallactic shift amount between match point in artwork, forms an image I and each put correspondence Parallactic shift amount.
Region to be fused is determined in the D1, is specially:According to coordinate map, in the short focus image of expansion, setting Some coordinate (i, j) so that
0<HPs(i,j)<1
Then record the coordinate value (i, j), be used as the starting point of integration region, it is desirable to the long width values Width of integration region and Height is no more than the correspondingly-sized of focal length image, so that it is determined that the position of lower integration region.
Beneficial effects of the present invention:
The present invention improves the local definition of infrared image by different focal image co-registration;
Parallel optical axis structure is that image registration reduces difficulty, by infinity image alignment, reduces matching search Dimension, sets up pyramid structure and improves search speed, and possibility is provided for image real time fusion;Pass through low point of different focal Resolution thermal imaging system obtains image there is provided the more preferable partial picture of definition, and this is a kind of cheap solution.
Brief description of the drawings
Fig. 1 is the workflow diagram of the present invention;
Fig. 2 is step B in Fig. 1 and step C workflow diagram.
Embodiment
With reference to embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not It is limited to this.
Embodiment
As shown in figure 1, a kind of thermal-induced imagery definition enhancing method, comprises the following steps:
A gathers thermal-induced imagery, and carries out enhancing contrast pretreatment;The infrared chart seems by two directional lights Axle construction and the thermal infrared imager acquisition with different focal, the thermal-induced imagery include short focus image and focal length image;Tool Body is:
A1 carries out smoothing denoising processing using median filtering algorithm or bilateral filtering algorithm to image, wherein on intermediate value filter Ripple is calculated as below:
Xo=Mid { Xi},i∈Z
Wherein Z is the window ranges of medium filtering, XiIt is the pixel value in window, XoIt is the output of medium filtering, Mid tables Show and seek median.
It is calculated as below on bilateral filtering:
The filter coefficient determined using the filter coefficient and picture element interpolation that are determined by geometric space distance is superimposed, and is calculated Weight coefficient:
Wherein, (i, j) is template center's coordinate, and (k, l) is the coordinate points in template,2 HesIt is the side of two wave filters Poor coefficient.I (i, j) is image pixel value.
According to weight Weight (i, j, k, l), pixel output I'(i is calculated, j)
So as to calculate view picture filtering image I'(i, j).Random noise and grain noise in removal infrared image, and compared with Good maintains image border.
A2 strengthens image high frequency detail using Butterworth sharpening filter, lifts image overall intensity.
High pass filter function is as follows:
Wherein D0For distance of the cut-off frequency away from origin.D (u, v) is defined as follows:
(u, v)=(M/2, N/2) is the center of frequency rectangle.
If fhp(x, y) represents the image after high-pass filtering, and f (x, y) represents artwork, it is considered to while sharpening image, also need The overall intensity of original image is lifted, then output image fout(x, y) should add low frequency component.
fout(x, y)=Af (x, y)+fhp(x,y)
Wherein, it is desirable to A>0.Following parameter combination, A=1.9, D are selected through experiment0=30.
As shown in Fig. 2 B expands short focus image, and focal length image is divided into low frequency and high frequency imaging;
B1 is according to the ratio between focal length of telephoto lens and short-focus lens, and such as telephoto lens 50mm focal lengths, short-focus lens 25mm is burnt Away from, then according to same ratio expansion short focus image, just twice of expansion here.So as to ensure the short focus image and length after to expand Burnt image has identical ratio scale in public domain.
B2 uses gauss low frequency filter, focal length image is divided into two images of high and low frequency, the target of filtering is to make The fog-level for obtaining low-frequency image is suitable with the expansion short focus image that (B1) is obtained.Identical fog-level is for low-frequency image Between preferably match, and high frequency imaging be used for image co-registration.
C is put down using DLT algorithms and translation mapping table, the infinity of the short focus image set up after expanding and original focal length image In face of neat mapping relations, it is specially:
Two camera lenses of C1 ensure that parallax range is as far as possible small according to parallel optical axis structure lay out in parallel, such as 5cm or so, The direction 1-2m sampled in Vertical camera lenses is placed and 20 spot lights is uniformly placed on one piece of surface plate, plate, and spot light pastes to be red Piece LED, is easy to sample match point.
The short focus image that C2 is expanded B1, it is uniform on image with original focal length image as one group of demarcation photo Find out 20 and assemble point coordinates, be P respectivelylAnd Ps, and record;
C3 is to PlAnd PsIt is normalized.Wherein Pl' and Ps' it is P respectivelyl、PsCoordinate after normalization.
Pl'=TlPl
Ps'=TrPs
Wherein TlAnd TsIt is normalization matrix.DlIt is the side of the preceding coordinate of normalization Difference, E (X) and E (Y) are the expectation of coordinate value before normalization, TsStructure and TlIt is similar.
C4PlAnd PsBetween corresponding relation be:
Pl*hPs=0
Wherein,wi' and wiFor 1.
Above formula is organized into Ah=0 homogeneous equation form.
By 20 couples of Pl' and Ps' match point coordinate bring above formula into, build homogeneous equation, using SVD singular value decompositions,
A=UDVT
H is V last row.H is organized into again 3*3 matrix form.
C5 carries out renormalization to coordinate, obtains directly description PlAnd PsThe H-matrix of relation.
Pl'=hPs'
TlPl=khTsPs
Wherein k is proportionality coefficient, it is ensured that3rd component of three-dimensional vector is 1, order
So as to obtain the homography matrix for describing short focus image and focal length image.That is Pl=HPs
C6 pinpoints P on the basis of the short focus image expanded in each 16*16 square framesThe list that (i, j) is solved using C5 Answering property matrix computations go out mapping point Pl(i, j), wherein i, j are 16 multiple.And fixed point processing is carried out to the coordinate, multiply With 64 rounds, discrete coordinate map is obtained.
In each 16*16 of C7 square frame, the coordinate map side of calculating obtained using the method and C4 of bilinear interpolation The mapping point in inframe portion.First with the first interpolation square frame four edges of equation below, mapping point is calculated,
F (k)=(f (1) * (64-k)+f (2) * k)/64
Then equation below interpolation square frame interior zone is utilized, mapping point is calculated.
F (x, y)=(f (1) * (64-x) * (64-y)+f (2) * (x) * (64-y)+f (3) * (64-x) * (y)+f (4) * x* Y)/128 wherein f (1), f (2), f (3), f (4) refers respectively to the mapping point of four vertex positions of square frame.In each side Inframe portion, completes above-mentioned Interpolation Process, so as to obtain complete coordinate map.
The vertical thermal imaging system optical axis directions of C8 arrange a plane (more than 10m) farther out, and binocular thermal imaging system shoots two figures Picture, using the obtained coordinate maps of C7 and bilinear interpolation algorithm, calculates the corresponding projective transformation image (PT of focal length image Image).
C9 finds the corresponding points of several groups of PT images and short focus expanded view picture, calculate its horizontal parallax average (coordinate value it Difference),
Offset=abs (Pl-HPs)
And record.Whole mapping table is translated using the parallax value, you can obtain the coordinate of plane at infinity alignment Mapping table.
D sets up image pyramid, calculates the horizontal direction parallax between image, specifically includes following steps:
D1 is according to gained mapping table in C9, in short focus expanded view picture, if some coordinate (i, j) so that
0<HPs(i,j)<1
The coordinate value (i, j) is then recorded, the starting point of integration region is used as.It is required that the long width values Width of integration region and Height is no more than the correspondingly-sized of focal length image.So that it is determined that the position of lower integration region.The integration region is contained in both The public domain of image.
The integration region that D2 determines D1 obtains B2 low-frequency image as image J, respectively to above-mentioned figure as image I Image sequence as setting up three layers of pyramid structure.The 1/2 of the upper strata a width of lower floor of pyramid length.
D3 travels through the pixel of I image top layers, sets up 3*3 characteristic block, the mapping table obtained according to C9, in J figures Its hunting zone is limited as upper horizontal unilateral direction is scanned for, and according to the priori of depth and parallax relation, is used by mistake The poor difference of two squares (SSD) minimum is used as measurement similarity indices
(i', j') is that coordinate points (i, j) draw its corresponding point coordinates on image J by inquiring about mapping table.So as to really Fixed a number of thick match point parallactic shift amount.
The thick match point that D4 is determined according to D3, using SSD minimums as similarity evaluation index, is transferred to lower floor and carries out surrounding The accurate of four points is compared, until it is determined that parallactic shift amount between match point in artwork, forms an I image and each put correspondence Parallactic shift amount.
The coordinate map that D5 obtains C9 is added with the disparity vector table that D4 is obtained, and is used as the obtained expansions of description B1 The matching relationship of short focus image and focal length image.
The short focus image that D6 will be enlarged by is as background, since the integration region starting point that D1 is obtained, using D5 obtain it is comprehensive Mapping table is closed, using bilinear interpolation algorithm, the obtained high frequency imagings of B2 is integrated into come, increases the figure of public domain definition Picture.
The HFS of focal length image is fused in the short focus image of expansion by E, specific as follows:
Obtained coordinate map and obtained disparity vector table are done into add operation, the short of the expansion for describing to obtain is used as The matching relationship of burnt image and original focal length image.
Since obtained integration region starting point, using obtained comprehensive mapping table, using bilinear interpolation algorithm, To the integration region that is fused to of high frequency imaging in, so as to add the definition of public domain.
Original focal length image described in text is the focal length image not pre-processed, and focal length image is after treatment Focal length image.
The present invention gathers infrared image first with the thermal infrared imager of different focal.Then utilize perpendicular to optical axis direction Outerplanar plate and two images to be matched, it is established that the coordinate map of two image infinitys alignment.Last infrared image By Image Fusion, the HFS of focal length image is fused in the short focus image of expansion.
Above-described embodiment is preferably embodiment, but embodiments of the present invention are not by the embodiment of the invention Limitation, other any Spirit Essences without departing from the present invention and the change made under principle, modification, replacement, combine, simplification, Equivalent substitute mode is should be, is included within protection scope of the present invention.

Claims (4)

1. a kind of thermal-induced imagery definition enhancing method, it is characterised in that comprise the following steps:
A gathers thermal-induced imagery, and carries out enhancing contrast pretreatment;The infrared chart seems by two parallel optical axis knots Structure and the thermal infrared imager acquisition with different focal, the thermal-induced imagery include short focus image and focal length image;
B expands short focus image, and focal length image is divided into low frequency and high frequency imaging;
The B specifically includes following steps:
B1 expands pretreated short focus image so that short after expansion according to telephoto lens and the focal length ratio of short-focus lens Burnt image has identical ratio scale with focal length image in public domain;
B2 uses gauss low frequency filter, and pretreated focal length image is divided into two images of high and low frequency, the mesh of filtering Mark is to make the fog-level of low-frequency image suitable with the short focus image that A1 is obtained;
C utilizes DLT algorithms and translation mapping table, the plane at infinity pair of the short focus image set up after expanding and original focal length image Neat mapping relations;
D sets up image pyramid, calculates the horizontal direction parallax between image;
D specifically includes following steps:
D1 is according to obtained coordinate map, in the short focus image of expansion, determines a region to be fused, obtains region to be fused Coordinate starting point and region to be fused size, i.e., the public domain between low-frequency image is obtained with B2;
Region to be fused is determined in the D1, is specially:According to coordinate map, in the short focus image of expansion, some is set Coordinate (i, j) so that
0<HPs(i,j)<1
Then record the coordinate value (i, j), be used as the starting point of integration region, it is desirable to the long width values Width of integration region and Height is no more than the correspondingly-sized of focal length image, so that it is determined that the position of lower integration region;
Region to be fused as image I, using low-frequency image as image J, on the basis of coordinate map, is calculated figure by D2 As the disparity vector of each points to be matched of I, a disparity vector table is formed;
The D2 specific steps include:
Three layers of pyramid as image I, using low-frequency image as image J, are set up in region to be fused by D21 to above-mentioned image respectively The image sequence of structure, described three layers golden word layer include upper, middle and lower layer, the upper strata pyramid length and a width of lower floor's pyramid Long and wide 1/2;
The pixel on D22 traversal I images upper strata, sets up 3*3 characteristic block, according to coordinate map, the level on image J Unilateral direction is scanned for, and limits its hunting zone according to the experience of depth and parallax relation, the use of square-error difference is SSD As measurement similarity indices,
<mrow> <mi>S</mi> <mi>S</mi> <mi>D</mi> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>m</mi> <mo>=</mo> <mo>-</mo> <mn>1</mn> </mrow> <mn>1</mn> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>n</mi> <mo>=</mo> <mo>-</mo> <mn>1</mn> </mrow> <mn>1</mn> </munderover> <msup> <mrow> <mo>(</mo> <mi>I</mi> <mo>(</mo> <mrow> <mi>i</mi> <mo>+</mo> <mi>m</mi> <mo>,</mo> <mi>j</mi> <mo>+</mo> <mi>n</mi> </mrow> <mo>)</mo> <mo>-</mo> <mi>J</mi> <mo>(</mo> <mrow> <msup> <mi>i</mi> <mo>&amp;prime;</mo> </msup> <mo>+</mo> <mi>m</mi> <mo>,</mo> <msup> <mi>j</mi> <mo>&amp;prime;</mo> </msup> <mo>+</mo> <mi>n</mi> </mrow> <mo>)</mo> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow>
(i', j ') is that coordinate points (i, j) draw its corresponding point coordinates on image J by inquiring about mapping table, so that it is determined that one The thick match point parallactic shift amount of fixed number amount;
D23, using SSD as similarity evaluation index, is directly transferred to lower floor and carries out surrounding four according to the thick match point of determination The accurate of point is compared, until it is determined that parallactic shift amount between match point in artwork, forms an image I and each put corresponding regard Poor offset;
The HFS of focal length image is fused in the short focus image of expansion by E.
2. Enhancement Method according to claim 1, it is characterised in that the A specifically includes following steps:
A1 carries out smoothing denoising processing using median filtering algorithm or bilateral filtering algorithm to image;
A2 strengthens image high frequency detail using Butterworth sharpening filter, lifts image overall intensity.
3. Enhancement Method according to claim 1, it is characterised in that the C specifically includes following steps:
Two camera lenses of C1 are according to parallel optical axis structure lay out in parallel, and the direction 1-2m sampled in Vertical camera lenses places one piece of plane Multiple spot lights are uniformly placed on plate, plate;
The short focus image that C2 is expanded B1, with original focal length image as one group of demarcation photo, uniformly finds out on image Multigroup matching point coordinates, and record;
C3 utilizes above-mentioned multigroup match point, and description short focus image and original focal length image are calculated using straight linear transfer algorithm Based on the outerplanar homography matrix of surface plate;
C4 calculates short focus image and original length on the basis of the short focus image expanded using homography matrix and bilinear interpolation algorithm The coordinate map of burnt image;
C5 removes surface plate described in C1, and multiple spot lights are then placed in the plane more than 10m, by calculating parallax, obtains The coordinate map of plane at infinity alignment.
4. Enhancement Method according to claim 3, it is characterised in that E specifically includes following steps:
Coordinate map and disparity vector table are done add operation and obtain comprehensive mapping table by E1, the short focus image expanded as description With the matching relationship of original focal length image;
The coordinate starting point in E2 regions to be fused starts, and using comprehensive mapping table and bilinear interpolation algorithm, high frequency imaging is merged Into region to be fused, increase the definition of public domain.
CN201410421814.2A 2014-08-25 2014-08-25 A kind of thermal-induced imagery definition enhancing method Expired - Fee Related CN104200450B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410421814.2A CN104200450B (en) 2014-08-25 2014-08-25 A kind of thermal-induced imagery definition enhancing method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410421814.2A CN104200450B (en) 2014-08-25 2014-08-25 A kind of thermal-induced imagery definition enhancing method

Publications (2)

Publication Number Publication Date
CN104200450A CN104200450A (en) 2014-12-10
CN104200450B true CN104200450B (en) 2017-08-25

Family

ID=52085736

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410421814.2A Expired - Fee Related CN104200450B (en) 2014-08-25 2014-08-25 A kind of thermal-induced imagery definition enhancing method

Country Status (1)

Country Link
CN (1) CN104200450B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017179295A1 (en) * 2016-04-13 2017-10-19 ソニー株式会社 Signal processing apparatus and image pickup apparatus
KR101795271B1 (en) * 2016-06-10 2017-11-07 현대자동차주식회사 Image Processing Apparatus and Method for Performing Pre-process for Clean Image
CN109937568B (en) * 2016-11-17 2021-03-09 索尼公司 Image processing apparatus, image processing method, and program
CN108401104B (en) * 2018-02-08 2020-11-13 浙江大学 Dual-focus camera digital zooming method based on frequency band repair and super-resolution
CN109285136B (en) * 2018-08-31 2021-06-08 清华-伯克利深圳学院筹备办公室 Multi-scale fusion method and device for images, storage medium and terminal
CN111242880B (en) * 2019-12-30 2023-05-02 广州市明美光电技术有限公司 Multi-depth-of-field image superposition method, equipment and medium for microscope
CN111818262B (en) 2020-07-08 2021-12-03 杭州萤石软件有限公司 Image reconstruction method and device
CN114283081B (en) * 2021-12-16 2022-11-08 合肥的卢深视科技有限公司 Depth recovery method based on pyramid acceleration, electronic device and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1484040A (en) * 2003-07-24 2004-03-24 上海交通大学 Image merging method based on filter group
EP2381417B1 (en) * 2010-04-23 2012-10-24 Flir Systems AB Infrared resolution and contrast enhancement with fusion
CN103632334A (en) * 2013-11-13 2014-03-12 华南理工大学 Infinite image alignment method based on parallel optical axis structure cameras
CN103776419A (en) * 2014-01-24 2014-05-07 华南理工大学 Binocular-vision distance measurement method capable of widening measurement range
CN103971352A (en) * 2014-04-18 2014-08-06 华南理工大学 Rapid image splicing method based on wide-angle lenses

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1484040A (en) * 2003-07-24 2004-03-24 上海交通大学 Image merging method based on filter group
EP2381417B1 (en) * 2010-04-23 2012-10-24 Flir Systems AB Infrared resolution and contrast enhancement with fusion
CN103632334A (en) * 2013-11-13 2014-03-12 华南理工大学 Infinite image alignment method based on parallel optical axis structure cameras
CN103776419A (en) * 2014-01-24 2014-05-07 华南理工大学 Binocular-vision distance measurement method capable of widening measurement range
CN103971352A (en) * 2014-04-18 2014-08-06 华南理工大学 Rapid image splicing method based on wide-angle lenses

Also Published As

Publication number Publication date
CN104200450A (en) 2014-12-10

Similar Documents

Publication Publication Date Title
CN104200450B (en) A kind of thermal-induced imagery definition enhancing method
CN106408524B (en) Depth image enhancement method based on two-dimensional image assistance
CN106780590B (en) Method and system for acquiring depth map
CN101356546B (en) Image high-resolution upgrading device, image high-resolution upgrading method image high-resolution upgrading system
CN108596975B (en) Stereo matching algorithm for weak texture region
CN109447930B (en) Wavelet domain light field full-focusing image generation algorithm
CN111145134B (en) Block effect-based microlens light field camera full-focus image generation algorithm
Namboodiri et al. On defocus, diffusion and depth estimation
CN104363369B (en) The image recovery method and device of a kind of light-field camera
CN105654547B (en) Three-dimensional rebuilding method
CN107169947B (en) Image fusion experimental method based on feature point positioning and edge detection
CN103106688A (en) Indoor three-dimensional scene rebuilding method based on double-layer rectification method
CN104240289A (en) Three-dimensional digitalization reconstruction method and system based on single camera
CN106023230B (en) A kind of dense matching method of suitable deformation pattern
US8867826B2 (en) Disparity estimation for misaligned stereo image pairs
Lee et al. Depth estimation from light field by accumulating binary maps based on foreground–background separation
CN107990846B (en) Active and passive combination depth information acquisition method based on single-frame structured light
CN105005964A (en) Video sequence image based method for rapidly generating panorama of geographic scene
CN109064505A (en) A kind of depth estimation method extracted based on sliding window tensor
CN108460795A (en) Inverted pyramid formula binocular solid matching process and system
Jiao et al. Joint image denoising and disparity estimation via stereo structure pca and noise-tolerant cost
CN110097634A (en) A kind of terrible imaging method of the three-dimensional of self-adapting multi-dimension
CN106023189A (en) Light field data depth reconstruction method based on matching optimization
CN103559701A (en) Two-dimensional single-view image depth estimation method based on DCT coefficient entropy
CN103873773B (en) Primary-auxiliary synergy double light path design-based omnidirectional imaging method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20170825

Termination date: 20210825

CF01 Termination of patent right due to non-payment of annual fee