CN104252704A - Total generalized variation-based infrared image multi-sensor super-resolution reconstruction method - Google Patents
Total generalized variation-based infrared image multi-sensor super-resolution reconstruction method Download PDFInfo
- Publication number
- CN104252704A CN104252704A CN201410477298.5A CN201410477298A CN104252704A CN 104252704 A CN104252704 A CN 104252704A CN 201410477298 A CN201410477298 A CN 201410477298A CN 104252704 A CN104252704 A CN 104252704A
- Authority
- CN
- China
- Prior art keywords
- infrared image
- resolution
- image
- value
- obtains
- 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.)
- Pending
Links
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses a total generalized variation-based infrared image multi-sensor super-resolution reconstruction method. The total generalized variation-based infrared image multi-sensor super-resolution reconstruction method mainly comprises the steps of projecting a low-resolution infrared image into the coordinate space of a high-resolution visible image, obtaining a sparse infrared image and solving a data item weighting coefficient according to the sparse infrared image; performing normalization processing on the sparse infrared image and obtaining a normalization infrared image; solving the marginal information of the high-resolution visible image through a phase equalization algorithm; constructing a data item by the data item weighting coefficient and the normalization infrared image; weighting a TGV regular term improved by a first-order gradient operator through the marginal information of the visible image and constructing a regular bound term; adding the data item and the regular bound term to construct an objective function, solving the objective function in an iterative mode through a primal-dual optimization algorithm with the normalization infrared image serving as an initial value and obtaining a reconstructed high-resolution infrared image. Experiments show that the quality of the image reconstructed by the total generalized variation-based infrared image multi-sensor super-resolution reconstruction method is high and the image is close to an original high-resolution infrared image.
Description
Technical field
The present invention relates to a kind of infrared image processing technology, be specifically related to a kind of High Resolution Visible Light image that utilizes and guide low resolution infrared image to carry out the method for super-resolution rebuilding, belong to infrared image super resolution technology field.
Background technology
The radiation characteristic of infrared image reflection scene, this characteristic can provide valuable information for multiple application, as military surveillance and Long-distance Control etc., but compared with visible images, infrared image edge is fuzzy and lack texture information, the infrared image limited spatial resolution directly obtained by infrared sensor, the information that thus extracting directly is enough from infrared image is more difficult.The spatial resolution improving infrared image is a problem demanding prompt solution.
In order to solve the problems referred to above that infrared image exists, people study from hardware aspect, attempt to solve the problems referred to above by the infrared focal plane device of the little pixel dimension of development high density, but under prior art conditions, the infrared focal plane device of the little pixel dimension of development high density also has certain difficulty, and development cost is higher, development efficiency is low.Given this, scientific research personnel explores from theory of algorithm aspect, is improved the resolution of infrared image by software.Software study aspect, a kind of promising method of tool utilizes signal processing technology to obtain high-definition picture by low-resolution image, claims this method to be super-resolution technique.Current super-resolution technique is mainly divided into three major types: 1) based on the super-resolution technique of interpolation, and the gray-scale value of neighborhood pixels is to produce the gray-scale value of interpolation pixel, obtains high-definition picture with this namely to utilize oneself to know.The high-definition picture of this reconstruction is second-rate at discontinuous places such as edges.2) based on the super-resolution technique of reconstruct, this technology, to the degenerative process modeling of image, utilizes low-resolution image and the process of reconstruction of specific image degradation model to high-definition picture to retrain, obtains high-definition picture.There is the problems such as priori is few, the effect improvement potentiality of reconstruction image are little in this technology.3) based on the super-resolution technique of study, this technical construction low resolution and high-definition picture Sample Storehouse, the inner link of low-resolution image and high-definition picture is obtained by learning sample storehouse, thus the super-resolution rebuilding process of guide image, obtain high-definition picture.In this technology, the training process computation complexity of Sample Storehouse is larger.Itself there is shortcoming in above traditional image super-resolution rebuilding method, and be mostly that the image utilizing same sensor to obtain is rebuild, but the image information that same sensor obtains is limited, more comprehensive available image information can not be provided, the quality defect of the high-resolution Thermo-imaging system therefore rebuild based on classic method.
Summary of the invention
For software aspect by the present situation of low resolution Infrared image reconstruction high-resolution Thermo-imaging system technology and deficiency, object of the present invention aims to provide a kind of infrared image super-resolution new method based on multi-sensor technology, the shortcoming of traditional images super-resolution technique is improved with this, and make the edge details of rebuilding the high-resolution Thermo-imaging system obtained clear, make up the defect that infrared image self is intrinsic to a certain extent.
There is very strong correlativity and complementarity with between scene visible images and infrared image, visible images edge details is clear, and Visible imaging system obtain image resolution ratio higher, these just infrared image be difficult to possess key character.Super-resolution technique aspect, regularization method obtains extensive concern because effectively trying to achieve optimization solution when it solves and rebuilds high-definition picture by low-resolution image.Basic thought of the present invention utilizes High Resolution Visible Light image to guide low resolution infrared image to carry out Super-resolution Reconstruction, and obtain the high-resolution Thermo-imaging system optimized in conjunction with regularization method.Based on this basic thought, the present invention proposes a kind of new method obtaining high-resolution Thermo-imaging system based on multi-sensor technology and image regulation super-resolution technique.
The present invention is in conjunction with the image of two kinds of different sensors acquisitions of same scene, and the technical scheme utilizing regularization method to obtain high-resolution Thermo-imaging system is: first utilize infrared image to carry out compose data items; Secondly extract visible images marginal information by phase equalization algorithm, utilize First-order Gradient operator to improve total GENERALIZED VARIATIONAL (TGV) regularization model, the TGV regular terms be improved; Then utilize the marginal information of visible images to the TGV regular terms weighting improved, obtain final canonical bound term; Finally adopt single order master-antithesis optimized algorithm to try to achieve optimum solution, optimum solution is the high-resolution Thermo-imaging system of optimized reconstruction.
Infrared image super-resolution method based on multi-sensor technology provided by the invention, particular content mainly comprises the following steps:
(1) pixel of low resolution infrared image is evenly diffused into the coordinate space of High Resolution Visible Light image, obtains sparse infrared image, and infrared image sparse thus obtains the weighting coefficient of data item;
(2) the sparse infrared image that step (1) obtains is normalized, obtains normalized infrared image;
(3) extract the edge of High Resolution Visible Light image according to phase equalization algorithm, obtain the high frequency edge information of High Resolution Visible Light image;
(4) the normalization infrared image compose data items that the data item weighting coefficient obtained by step (1) and step (2) obtain;
(5) the TGV regular terms weighting that marginal information step (3) obtained is improved First-order Gradient operator, obtains canonical bound term;
(6) the canonical bound term that data item step (4) obtained and step (5) obtain is added the objective function obtaining high-resolution Thermo-imaging system as the present invention, utilize normalization infrared image that step (2) obtains as initial value, adopt master-antithesis optimized algorithm iterative objective function, obtain high-resolution Thermo-imaging system.
In technique scheme of the present invention, the specific rules that step (1) obtains data item weighting coefficient by sparse infrared image is: the coordinate points coefficient value not having pixel to cover at sparse infrared image is 0, otherwise is 1.
In technique scheme of the present invention, step (2) takes following method to calculate normalized infrared image:
1) obtain the set that sparse infrared image grey scale pixel value that step (1) obtains is greater than 0, then obtain minimum value and the maximal value of grey scale pixel value in this set;
2) normalization infrared image is tried to achieve according to the minimum value of above-mentioned grey scale pixel value, maximal value and sparse infrared image.
In technique scheme of the present invention, step (3) utilizes phase equalization algorithm to ask visible images marginal information
computing formula as follows:
In above formula
, represent angle;
represent the phase value of signal in x place n-th Fourier components;
that above formula is existed
xthe weighted mean of its Fourier each component local phase angle when maximal value is got at place;
represent that signal exists
xthe value of the phase equalization at place;
for frequency propagates weighted volumes;
for signal is at the range value of x place n-th Fourier components;
be one avoid denominator be 0 very little constant;
for noise threshold;
being a mathematical operation, when desired value is that timing net result takes from body, otherwise is 0.
In technique scheme of the present invention, the expression formula of step (4) institute compose data items is:
In above formula
for the high-resolution Thermo-imaging system operation result after each iterative computation,
for the normalization infrared image that step (2) obtains,
for the data item weighting coefficient that step (1) obtains, C is the coordinate space of High Resolution Visible Light image.
In technique scheme of the present invention, the method for step (5) structure canonical bound term is as follows:
1) First-order Gradient operator is utilized to improve TGV regular terms.First-order Gradient operator and TGV regular terms combine by the present invention, and gained expression formula is as follows:
Constant in above formula
,
with
for weight parameter,
for the symmetric matrix in TGV regularization algorithm, the expression formula of above-described First-order Gradient operator is as follows:
The expression formula of TGV regular terms is as follows:
2) marginal information that step (3) is tried to achieve is utilized
the TGV regular terms improved is weighted.Gained expression formula is:
Above formula is final canonical bound term.
In technique scheme of the present invention, the process that step (6) obtains objective function and iterative comprises following content:
1) the canonical bound term that data item step (4) obtained and step (5) obtain is added as acquisition high-resolution Thermo-imaging system
objective function:
2) single order master-antithesis optimized algorithm iterative computation is adopted to objective function, above formula is converted into:
Wherein
,
be the dual variable in master-antithesis optimized algorithm, the set of its place is as follows respectively:
Wherein
for the pixel coordinate in image.
The present invention adopts single order master-antithesis optimized algorithm to solve objective function in conjunction with gradient descent method, master variable elected as during concrete iterative computation
with
, step-length
with
be taken as the constant being greater than 0, during first time iteration,
be taken as
, get
,
,
,
,
be 0, comprise the following steps:
The first, dual variable is upgraded by gradient ascent iterations:
In above formula
with
for the intermediate result of iterative computation;
The second, master variable is upgraded by Gradient Descent iteration:
In above formula
,
for step-length, its value changes with the difference of input picture;
3rd, master variable is optimized further, and computing formula is as follows:
In above formula
value all specifically upgrade when each iteration;
4th, the present invention sets the iterations upper limit, the Output rusults when the result of twice interative computation in front and back meets the following conditions:
In above formula
for error threshold, the resolution of required high-resolution Thermo-imaging system is
×
.If do not meet above-mentioned condition, then the first to three steps of the as above iteration optimization algorithms of circulating, until upper limit number of times, then export the result of calculation of high-resolution Thermo-imaging system.
The objective function of the present invention's algorithm used is added by data item and canonical bound term and forms, wherein data item is mainly used in the similarity ensureing the high-resolution Thermo-imaging system after rebuilding and original low resolution infrared image, can limit the deviation of reconstructed results and original low-resolution image.Canonical bound term utilizes the approximate solution space of the prior-constrained condition of image to the high-resolution Thermo-imaging system rebuild to retrain, thus obtains the optimum solution in approximate solution space.
The present invention proposes a kind of regularization model First-order Gradient operator and TGV regularization method combined.TGV regularization method effectively can retain the marginal information of super-resolution rebuilding image, alleviates staircase effect simultaneously.But infrared imaging system can be mixed into multiple noise in imaging process, as system noise, thermonoise and random noise etc.; On the other hand, between imaged scene there is heat interchange in the moment, and cause object heat radiation degree relative equilibrium, make become infrared image contrast low, edge is comparatively fuzzy.Therefore when TGV regularization method carries out super-resolution rebuilding to infrared image, the image border of acquisition is clear not, and there is noise.The present invention considers that First-order Gradient operator can suppress low-frequency noise and sharpening image edge, and in actual applications, sharp filtering can strengthen the edge of details by fuzzy or soft image.Therefore the present invention utilizes First-order Gradient operator to improve TGV regularization model, obtains a kind of follow-on TGV regularization model, and this model can the solution space of operative constraint infrared image super-resolution, obtains the high-resolution Thermo-imaging system that quality is higher.
The present invention utilizes phase equalization algorithm to extract the edge of visible images, and the visible images marginal information extracted with this is supplemented infrared image.The present invention adopts phase equalization algorithm to overcome the defect of conventional edge extraction algorithm.Though traditional edge extracting method can detection and positioning step edge feature, have ignored line edge, roof and the marginal information between step edge and line edge, therefore, the edge obtained is the defect such as jagged, burr often.By contrast, the important feature of phase equalization only simply unanimously finds unique point by phase place at Fourier transform domain.The characteristic types such as various line, step, roof can make high phase equalization point occur, thus obtain extracting result.This algorithm has advantages such as the robustnesss that illumination condition and contrast change, and in the very weak place of object edge feature, also can detect image border preferably, be a high performance Image Feature Detection Algorithms.Therefore the high efficiency extraction of phase equalization algorithm realization visible images marginal information is utilized herein.
The present invention utilizes High Resolution Visible Light image to guide low resolution infrared image to carry out super-resolution rebuilding, the marginal information of visible images is utilized to carry out rational weighting to modified TGV model, the important high-frequency information that infrared image lacks by the model after weighting introduces reconstructed results, effectively make use of the abundant accurate advantage again of High Resolution Visible Light image edge information, obtain the more accurate infrared image of details.
The present invention is after compose data items and canonical bound term, two are added and obtain objective function, then using normalized infrared image as initial value, adopt master-antithesis optimized algorithm in conjunction with gradient descent method iterative objective function, try to achieve high-resolution Thermo-imaging system with this.
A kind of method based on multi-sensor technology acquisition high-resolution Thermo-imaging system that the present invention proposes can bring following effect:
The present invention's infrared image carrys out compose data items; Phase equalization algorithm is adopted to extract the edge of visible images, First-order Gradient operator is combined with TGV regularization method and obtains a kind of TGV model of improvement, then utilize the visible images marginal information extracted to the TGV model-weight improved, obtain canonical bound term; Finally try to achieve optimum solution by master-antithesis optimized algorithm in conjunction with gradient descent method.Experiment shows that this algorithm has good performance, and the high-resolution infrared image quality reconstructed is higher;
The present invention is in order to the marginal information of effective acquisition visible images, phase equalization algorithm is utilized to extract the edge of High Resolution Visible Light image, this algorithm is not by the impact that image irradiation and contrast change, can obtain complete and reliable image border when brightness changes, be a high performance edge detection algorithm.Therefore the present invention utilizes phase equalization algorithm to extract the edge of visible images, achieve High Resolution Visible Light information rationally and extract efficiently, experiment proves, after the visible light image information obtained with phase equalization algorithm joins and rebuilds image, make the edge details of rebuilding high-resolution Thermo-imaging system more clear;
TGV regular terms combines with First-order Gradient operator by the present invention, proposes a kind of TGV regularization model of improvement.This model can effectively suppress to rebuild the random noise of high-resolution Thermo-imaging system, simultaneously reserved high-frequency edge and alleviate staircase effect.Simultaneously, the marginal information that phase equalization algorithm extracts by the present invention is to the TGV model-weight improved, the information of visible images is rationally introduced infrared image by the regularization model after weighting, the comparatively accurate and comprehensive high-resolution Thermo-imaging system of the information that can obtain.Experimental result shows, rebuilds the high-resolution Thermo-imaging system obtained and has good visual effect.
Accompanying drawing explanation
Fig. 1 is the schematic process flow diagram of the inventive method;
Fig. 2 is the schematic diagram that low resolution infrared image projects High Resolution Visible Light image coordinate space;
Fig. 3 is that the present invention tests high-definition picture used, and wherein (a) is visible images, and (b) is infrared image;
Fig. 4 is that the present invention tests low resolution infrared image input figure used;
Fig. 5 is the high-resolution Thermo-imaging system of experimental reconstruction of the present invention.
Embodiment
Below in conjunction with specific experiment, the present invention is described in detail, but should not be understood as any restriction to scope.
Invention has been specific experiment to verify the validity of put forward algorithm.Fig. 3 experiment figure is visible images and the infrared image in house, and High Resolution Visible Light image and high-resolution Thermo-imaging system are same scene image, and resolution is 320 × 240.
The first step, carries out high-resolution infrared image the process that degrades, i.e. 2 times of down-samplings.Adopt the method for gaussian pyramid down-sampling to obtain in the present invention's experiment low resolution infrared image that resolution is 160 × 120
, as shown in Figure 4.
Second step, the low resolution infrared image that the first step is obtained
be diffused into the coordinate space of High Resolution Visible Light image, obtain sparse infrared image
, schematic diagram as shown in Figure 2.Projection process can be described by following formula:
（1）
On the right of above formula in bracket
=1,2 ... 160,
=1,2 ... 120, in the bracket of the left side
=1,2 ... 320,
=1,2 ... 240, and
,
, be that the pixel of the low resolution infrared image of 160 × 120 is evenly diffused into the coordinate space that resolution is the High Resolution Visible Light image of 320 × 240 by resolution.Calculate weighted operator thus
,
, at sparse infrared image
the coordinate points place not having pixel to cover
, otherwise
.
3rd step, to the infrared image that second step obtains
be normalized.Comprise the following steps, wherein
=1,2 ... 320,
=1,2 ... 240:
1) pixel grey scale value set is obtained
in minimum value
and maximal value
, account form is as follows:
（2）
2) infrared image after preliminary normalization is calculated
, computing formula is as follows:
（3）
3) normalization infrared image is tried to achieve by with under type:
（4）
be normalized infrared image.
4th step, calculates the edge of High Resolution Visible Light image according to phase equalization algorithm
, computing formula is as follows:
（5）
（6）
In above formula
, represent angle;
represent the phase value of signal in x place n-th Fourier components;
that above formula is existed
xthe weighted mean of its Fourier each component local phase angle when maximal value is got at place;
represent that signal exists
xthe value of the phase equalization at place;
for frequency propagates weighted volumes;
for signal is at the range value of x place n-th Fourier components;
be one avoid denominator be 0 very little constant;
for noise threshold;
being a mathematical operation, when desired value is that timing net result takes from body, otherwise is 0.
5th step, the present invention adopts master-antithesis optimized algorithm to go out optimum solution in conjunction with gradient descent method iterative, and dual variable is expressed as
with
, master variable is elected as
with
.During first time iteration,
be taken as that the 3rd step obtains
, get
,
,
,
,
be 0, constant in experiment
get 0.333,
get 0.5,
get 0.08,
get 0.6,
get 3, comprise following calculation procedure:
1) dual variable is upgraded by gradient ascent iterations, and computing formula is as follows:
（7）
In above formula
with
for the intermediate result of iterative computation;
2) master variable is upgraded by Gradient Descent iteration, is obtained by second step
, the 3rd step obtains
, the 4th step obtains
, being risen by gradient is calculated
with
substitute into following formula:
（8）
In above formula
,
for step-length, its value changes with the difference of input picture;
3) master variable is optimized further, and computing formula is as follows:
（9）
In above formula
value all specifically upgrade when each iteration, in experiment will
initial value be set to 0.77;
4) in the present invention's experiment, the iterations upper limit is set to 1000 times, exports reconstructed results when the result of twice interative computation in front and back meets the following conditions:
（10）
In experiment
get 0.1, if do not meet above-mentioned condition, then 1 of the as above iteration optimization algorithms of circulating) ~ 3) step until upper limit number of times, then export the result of calculation of high-resolution Thermo-imaging system.
Experiment shows that algorithm of the present invention has good performance, and picture noise is less, is really true to life.As shown in Figure 5, the combination of High Resolution Visible Light image and low resolution infrared image, makes the marginal information adding High Resolution Visible Light image in reconstructed results, and the high-resolution Thermo-imaging system quality reconstructed is better, edge details is clear, close to original high resolution infrared image.
Claims (9)
1., based on a method for the infrared image multisensor super-resolution rebuilding of total GENERALIZED VARIATIONAL, its feature mainly comprises the following steps:
(1) pixel of low resolution infrared image is evenly diffused into the coordinate space of High Resolution Visible Light image, obtains sparse infrared image, and infrared image sparse thus obtains the weighting coefficient of data item;
(2) the sparse infrared image that step (1) obtains is normalized, obtains normalized infrared image;
(3) extract the edge of High Resolution Visible Light image according to phase equalization algorithm, obtain the high frequency edge information of High Resolution Visible Light image;
(4) the normalization infrared image compose data items that the data item weighting coefficient obtained by step (1) and step (2) obtain;
(5) total GENERALIZED VARIATIONAL (TGV) the regular terms weighting that marginal information step (3) obtained is improved First-order Gradient operator, obtains canonical bound term;
(6) the canonical bound term that data item step (4) obtained and step (5) obtain is added the objective function as obtaining high-resolution Thermo-imaging system, utilize normalization infrared image that step (2) obtains as initial value, adopt master-antithesis optimized algorithm iterative objective function, obtain high-resolution Thermo-imaging system.
2. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 1, it is characterized in that, the specific rules that step (1) obtains data item weighting coefficient by sparse infrared image is: the coordinate points coefficient value not having pixel to cover at sparse infrared image is 0, otherwise is 1.
3. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 1, it is characterized in that, step (2) takes following method to calculate normalized infrared image:
1) obtain the set that sparse infrared image grey scale pixel value that step (1) obtains is greater than 0, then obtain minimum value and the maximal value of grey scale pixel value in this set;
2) normalization infrared image is tried to achieve according to the minimum value of above-mentioned grey scale pixel value, maximal value and sparse infrared image.
4. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 1, is characterized in that, the computing formula that step (3) extracts visible images edge according to phase equalization algorithm is as follows:
θ ∈ [0, π] in above formula, represents angle; φ
n(x, θ) represents the phase value of signal in x place n-th Fourier components;
it is the weighted mean of its Fourier each component local phase angle when making above formula get maximal value at x place; P represents the value of signal in the phase equalization at x place; W (x, θ) is frequency propagation weighted volumes; A
n(x, θ) is for signal is at the range value of x place n-th Fourier components; ε be one avoid denominator be 0 very little constant; T is noise threshold;
being a mathematical operation, when desired value is that timing net result takes from body, otherwise is 0.
5. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 1, is characterized in that, the expression formula of step (4) institute compose data items is:
In above formula, u is the high-resolution Thermo-imaging system operation result after each iterative computation, I
nfor the normalization infrared image that step (2) obtains, ω is the data item weighting coefficient that step (1) obtains, and C is the coordinate space of High Resolution Visible Light image.
6. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 1, is characterized in that, the method for step (5) structure canonical bound term is as follows:
1) utilize First-order Gradient operator to improve TGV regular terms, gained expression formula is as follows:
Constant α in above formula
2, α
1and α
0for weight parameter, v is the symmetric matrix in TGV regularization algorithm;
2) the marginal information P utilizing step (3) to try to achieve is weighted the TGV regular terms improved, and gained expression formula is:
Above formula is final canonical bound term.
7. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 1, is characterized in that, step (6) described acquisition high-resolution Thermo-imaging system I
hthe expression formula of objective function be:
Above formula is added by data item and canonical bound term and forms.
8. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 7, is characterized in that, adopts single order master-antithesis optimized algorithm iterative computation, objective function is converted into following expression formula to objective function:
Wherein p, q are the bivariate in master-antithesis optimized algorithm, and the set of its place is as follows respectively:
9. the infrared image multisensor super resolution ratio reconstruction method based on total GENERALIZED VARIATIONAL according to claim 8, is characterized in that, adopts during single order master-antithesis optimized algorithm iterative computation concrete in conjunction with gradient descent method and elects master variable as u and v, step-length θ
pand θ
qfor being greater than the constant of 0, during first time iteration, u is taken as I
n, get v, p, q, u
0, v
0be 0, comprise the following steps:
1) dual variable is upgraded by gradient ascent iterations:
U in above formula
0and v
0for the intermediate result of iterative computation;
2) master variable is upgraded by Gradient Descent iteration:
K in above formula
u, k
vfor step-length, its value changes with the difference of input picture;
3) master variable is optimized further, and computing formula is as follows:
In above formula, the value of μ all specifically upgrades when each iteration;
4) the present invention sets the iterations upper limit, exports result of calculation when the result of twice interative computation in front and back meets the following conditions:
In above formula, e is error threshold, the resolution of required high-resolution Thermo-imaging system is M × N, if do not meet above-mentioned condition, then 1 of the as above iteration optimization algorithms of circulating) ~ 3) step until upper limit number of times, then export the result of calculation of high-resolution Thermo-imaging system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410477298.5A CN104252704A (en) | 2014-09-18 | 2014-09-18 | Total generalized variation-based infrared image multi-sensor super-resolution reconstruction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410477298.5A CN104252704A (en) | 2014-09-18 | 2014-09-18 | Total generalized variation-based infrared image multi-sensor super-resolution reconstruction method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104252704A true CN104252704A (en) | 2014-12-31 |
Family
ID=52187572
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410477298.5A Pending CN104252704A (en) | 2014-09-18 | 2014-09-18 | Total generalized variation-based infrared image multi-sensor super-resolution reconstruction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104252704A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616267A (en) * | 2015-02-17 | 2015-05-13 | 海南大学 | Method of rebuilding under-sampled image based on minimal second-order total generalized variation |
| CN105118038A (en) * | 2015-09-11 | 2015-12-02 | 哈尔滨工业大学（威海） | High-definition image reconstruction method for photoacoustic microscope based on second order total generalized variation |
| CN105405095A (en) * | 2015-09-28 | 2016-03-16 | 高新兴科技集团股份有限公司 | Image super-resolution reconstruction method based DOG operator |
| CN109255773A (en) * | 2018-09-13 | 2019-01-22 | 武汉大学 | Different resolution ratio based on full variation is infrared with visible light image fusion method and system |
| CN110063753A (en) * | 2019-03-18 | 2019-07-30 | 深圳蓝韵医学影像有限公司 | The method and apparatus of fetus head circumference is measured in ultrasound image |
| CN112132753A (en) * | 2020-11-06 | 2020-12-25 | 湖南大学 | Infrared image super-resolution method and system for multi-scale structure guide image |
| CN112200718A (en) * | 2020-09-18 | 2021-01-08 | 郑州航空工业管理学院 | Infrared image super-resolution method based on NCSR and multiple sensors |
-
2014
- 2014-09-18 CN CN201410477298.5A patent/CN104252704A/en active Pending
Non-Patent Citations (5)
| Title |
|---|
| PETER KOVESI: "Edges Are Not Just Steps", 《ACCV2002: THE 5TH ASIAN CONFERENCE ON COMPUTER VISION》 * |
| PETER KOVESI: "Phase Congruency Detects Corners and Edges", 《PROC. VIITH DIGITAL IMAGE COMPUTING: TECHNIQUES AND APPLICATIONS》 * |
| WEI WU等: "Learning-based super resolution using kernel partial least squares", 《IMAGE AND VISION COMPUTING》 * |
| 卢思军: "立体匹配关键技术研究", 《中国博士学位论文全文数据库信息科技辑（月刊）》 * |
| 邸维巍等: "彩色图约束的二阶广义总变分深度图超分辨率重建", 《中国图象图形学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104616267A (en) * | 2015-02-17 | 2015-05-13 | 海南大学 | Method of rebuilding under-sampled image based on minimal second-order total generalized variation |
| CN105118038A (en) * | 2015-09-11 | 2015-12-02 | 哈尔滨工业大学（威海） | High-definition image reconstruction method for photoacoustic microscope based on second order total generalized variation |
| CN105405095A (en) * | 2015-09-28 | 2016-03-16 | 高新兴科技集团股份有限公司 | Image super-resolution reconstruction method based DOG operator |
| CN109255773A (en) * | 2018-09-13 | 2019-01-22 | 武汉大学 | Different resolution ratio based on full variation is infrared with visible light image fusion method and system |
| CN109255773B (en) * | 2018-09-13 | 2021-05-04 | 武汉大学 | Different-resolution infrared and visible light image fusion method and system based on total variation |
| CN110063753A (en) * | 2019-03-18 | 2019-07-30 | 深圳蓝韵医学影像有限公司 | The method and apparatus of fetus head circumference is measured in ultrasound image |
| CN110063753B (en) * | 2019-03-18 | 2022-03-22 | 深圳蓝韵医学影像有限公司 | Method and device for measuring head circumference of fetus in ultrasonic image |
| CN112200718A (en) * | 2020-09-18 | 2021-01-08 | 郑州航空工业管理学院 | Infrared image super-resolution method based on NCSR and multiple sensors |
| CN112132753A (en) * | 2020-11-06 | 2020-12-25 | 湖南大学 | Infrared image super-resolution method and system for multi-scale structure guide image |
| CN112132753B (en) * | 2020-11-06 | 2022-04-05 | 湖南大学 | Infrared image super-resolution method and system for multi-scale structure guide image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104252704A (en) | Total generalized variation-based infrared image multi-sensor super-resolution reconstruction method | |
| CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
| CN110021037B (en) | Image non-rigid registration method and system based on generation countermeasure network | |
| CN103310453B (en) | A kind of fast image registration method based on subimage Corner Feature | |
| CN104008539B (en) | Image super-resolution rebuilding method based on multiscale geometric analysis | |
| CN104408700A (en) | Morphology and PCA (principal component analysis) based contourlet fusion method for infrared and visible light images | |
| CN107358576A (en) | Depth map super resolution ratio reconstruction method based on convolutional neural networks | |
| CN105046672A (en) | Method for image super-resolution reconstruction | |
| CN106228512A (en) | Based on learning rate adaptive convolutional neural networks image super-resolution rebuilding method | |
| CN106952228A (en) | The super resolution ratio reconstruction method of single image based on the non local self-similarity of image | |
| CN102800076B (en) | Image super-resolution reconstruction method based on double-dictionary learning | |
| CN104835130A (en) | Multi-exposure image fusion method | |
| CN106339996B (en) | A kind of Image Blind deblurring method based on super Laplace prior | |
| CN103150713A (en) | Image super-resolution method of utilizing image block classification sparse representation and self-adaptive aggregation | |
| CN103020909A (en) | Single-image super-resolution method based on multi-scale structural self-similarity and compressive sensing | |
| CN105139339A (en) | Polarization image super-resolution reconstruction method based on multi-level filtering and sample matching | |
| CN102163329A (en) | Super-resolution reconstruction method of single-width infrared image based on scale analogy | |
| CN103971354A (en) | Method for reconstructing low-resolution infrared image into high-resolution infrared image | |
| CN106886978A (en) | A kind of super resolution ratio reconstruction method of image | |
| CN102184532A (en) | Single scale based medical image edge detection | |
| CN109146797A (en) | A kind of impulsive noise ancient book image inpainting method sparse based on Lp pseudonorm and overlapping group | |
| CN104899835A (en) | Super-resolution processing method for image based on blind fuzzy estimation and anchoring space mapping | |
| CN105550989A (en) | Image super-resolution method based on nonlocal Gaussian process regression | |
| CN104657951A (en) | Multiplicative noise removal method for image | |
| Luo et al. | IMNN-LWEC: A novel infrared small target detection based on spatial–temporal tensor model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20141231 |