CN101146233A - A computing and image correction method for light source color - Google Patents
A computing and image correction method for light source color Download PDFInfo
- Publication number
- CN101146233A CN101146233A CNA2007101332729A CN200710133272A CN101146233A CN 101146233 A CN101146233 A CN 101146233A CN A2007101332729 A CNA2007101332729 A CN A2007101332729A CN 200710133272 A CN200710133272 A CN 200710133272A CN 101146233 A CN101146233 A CN 101146233A
- Authority
- CN
- China
- Prior art keywords
- pixel
- light source
- color
- point
- 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.)
- Pending
Links
Images
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
A light source color calculation and image correction method is provided, which relates to a calculation and correction method of a light source during the identification and analysis of color images. The method comprises the following steps: normalizing the colority of each pixel in a color image; rapidly extracting a highlight area by a voting method based on the change gradient of pixels by using pure diffuse reflection or less mirror reflection with respect to the surrounding pixels, that is, the property of large similarity degree of the pixels; deducing the linear relationship between inverse intensity and colority and the representation and correction method of the light source color based on a simplified bidirectional reflectance function illumination model, so as to respectively project the colority values of the three channels of the pixels in the highlight area to spaces defined by the inverse intensity and the colority and thus to transform a three-dimensional space into a two-dimensional space; linearly matching the points in the inverse intensity space by means of parameter approximation to obtain a colority axis intercept (that is light source color value); and performing the color correction on the normalized image based on the light source color value and then reverting to an image under standard white light conditions.
Description
Technical field
The present invention relates to a kind of calculating and method for correcting image of light source colour, relate in particular to the identification of coloured image and the calculating and the bearing calibration of the light source in the analysis, to keep the constant color of image.Belong to the Image Information Processing field.
Background technology
The performance of object on imaging system determines aspect three: the sensor performance of illumination, object surfaces attribute and imaging system.Because the color of light source under natural scene, generally is not pure white particularly, therefore when solid coupling, object identification, must consider the color of light source.In addition, the illumination variation of shooting environmental can cause image internal object change in color, in order to guarantee the reliability of analysis result, calculating and the correction of image being carried out light source colour are necessary links, have very important significance for visual identity, industrial detection.
Can be divided into based on statistics with based on two big classes of physics at the light source colour computational methods.
Utilize the relation of the probability statistics knowledge of light source and surface reflectivity and color to realize the calculating of color based on the light source colour computational methods of statistics.Representative Retinex algorithm (Land E H, McCann J J.Lightness and retinex theory.Journal of Optics Society of America, 1971,61 (1): 1-11) discontinuity of hypothesis color of object surface is greater than the discontinuity of illumination color.Utilization logarithm calculus of differences extracts the surface color discontinuity, and the discontinuity of place to go light source.Then, utilization integral exponential computing reconstructing surface color.Retinex algorithm and improvement algorithm on this basis (Weiss Y.Derivingintrinsic images from image sequences.IEEE Internat ional Conference onComputer Vision, 2003, subject matter 2:68-75) is: correctly find the discontinuity of color of object surface, and up to the present this is still a difficult problem.Shortcoming based on the light source colour computational methods of adding up is that the requirement body surface has multiple color.In addition, most methods hypothesis body surface has only diffuse reflection, and is subjected to the influence of noise easily based on the algorithm of statistics.
Along with the development of computer graphics, the researcher begins to study on certain illumination and reflection model.Can handle the situation on multicolour and complex texture surface well based on the method for physics, therefore become the main flow direction of research.Object color is made up of two parts: the body reflecting component of the reflecting component of body surface and object.The surface reflection component has reflected the colouring information of light source mainly by the reflecting attribute decision of light source.What the body reflection reflected is the colouring information of object itself.T type light source colour computational methods (Klinker G J, Shafer S A, Kanade T.Themeasurement of highlights in color images.International Journal of ComputerVision, 1990,2:7-32) calculate light source colour at RGB three dimensions by a solid color surface.At three-dimensional rgb space, the pixel family of 2 reflecting components forms a T type, can obtain light source colour by extraction and decomposition to the T type.But the image that obtains in the actual scene, because the influence of noise, the extraction of T type is very difficult.And three-dimensional amount of calculation is very big.For fear of three-dimensional complex calculation, document (Lee H C.Method for computing the scene-illuminant from specular highlights.Journalof Optics Society of America A, 1986,3 (10): 1694-1699) proposed a kind of computational methods of utilizing many colors highlight area to carry out light source colour.But this algorithm requires to utilize image partition method that high light is partly cut apart, and be not suitable for the texture complex surface, and the straight-line intersection of two similar colors is also inapplicable for the object that has only a kind of color with respect to the noise sensitivity.On this basis, document (Lehmann T M, PalmC.Color line search for illuminant estimation in real-world scene.Journalof Optics Society of America A, 2001,18 (11): 2679-2691) proposed a kind of more stable algorithm.But image needs the high light of multiple color to be existed, and to require the surface color in each highlight area be the same, can't handle the complex texture surface condition.Contrary Strength Space is calculated method (the TanR T of light source colour, Nishino K, Ikeuchi K.Illumination Chrmaticity estimation usinginverse-intensity chromaticity space.IEEE Computer Society Conference onComputer Vision and Pattern Recognition, 2003,673-680.) need not to carry out image segmentation, and to highlight area number and color without limits.But the method is too complicated on highlight area detection and Hough SPATIAL CALCULATION straight-line intersection, causes the processing time lengthening.Generally speaking, the defective of current light source computational methods based on physics is that most of algorithms also need to carry out image segmentation, causes speed slow, and the light source computational accuracy depends on the precision of cutting apart; Restricted to image texture, for owing texture and complex texture object precision is very low; The three-dimensional color space calculation of complex, length consuming time.
Summary of the invention
Technical problem: the objective of the invention is at the deficiencies in the prior art: speed is slow, to problems such as the restriction of texture, calculation of complex, a kind of image segmentation, to the object surfaces texture without limits and carry out a kind of light source colour that light source colour calculates at two-dimensional space and calculate and method for correcting image of need not is provided, and with the color of image rectification under the standard white light.
Technical scheme: from BRDF (the Bi-directional Reflectance DistributionFunction that simplifies, the bidirectional reflectance function) illumination model sets out, highlight area during employing voting mechanism rapid extraction is published picture and looked like, contrary Strength Space will be projected to then behind the image normalization, utilize pixel colourity and the linear relationship of contrary intensity in contrary Strength Space, carry out fitting a straight line by the parameter approach method, calculate the color value of light source.The light source colour that last basis calculates carries out color correction to image.
Utilize the present invention to carry out that light source colour calculates and method for correcting image may further comprise the steps:
Steps A. the colourity to each pixel in the coloured image is carried out normalization: the relation table of light source colour Г and color of object surface Λ is shown I (x)=m
d(x) Λ (x)+m
s(x) Г, the color value of each the pixel x among I (x) the presentation video I wherein, m
d(x) remarked pixel point x place object surfaces diffuse reflectance, A (x) remarked pixel point x place object surfaces color, m
s(x) remarked pixel point x place object surfaces specularity factor, Г represents light source colour, and the color of each pixel x in the image I is carried out normalization: σ (x)=I (x)/∑ I
c(x), wherein σ (x) is the colourity of pixel x after the normalization.∑ I
c(x) three passage color values of remarked pixel x sum; I
cRemarked pixel point x is at the color value of c Color Channel;
Step B. utilizes the variation of the pixel of pure diffuse reflection or a small amount of direct reflection and surrounding pixel point mild, be the big character of similarity of pixel, go out highlight area by the voting method rapid extraction: for the surf zone of same color, the pixel of pure diffuse reflection or a small amount of direct reflection and the variation of surrounding pixel point are mild, the similarity of pixel is very big, all pixels is carried out high light detect adjacent three pixel (x-1 in the same color region, x, x+1) the pixel color difference value is:
Λ (x-1 wherein, x) be the color distortion value of two neighbor x-1 and x, note ε=| Λ (x, x+1)-Λ (x-1, x) |, determine by voting mechanism whether a pixel is the diffuse reflection point, think so when ε＜α, be pixel x-1, x and x+1 are for throwing a ticket respectively, and α is the threshold value that is provided with in advance here; After having inspected all pixels, poll is exactly the diffuse reflection point greater than the point of preset value λ, and poll is exactly a specular reflection point less than the point of preset value; Each pixel will be checked 6 times in this algorithm, is considered to marginal point when the color of neighbor pixel and light intensity differ greatly in addition, also is excluded outside highlight area;
The BRDF illumination model of step C. from simplifying, derive contrary intensity and the linear relationship of colourity and the expression and the bearing calibration of light source colour, the chromatic value of three passages of the pixel in the highlight area is projected respectively by in contrary intensity and the space that colourity is opened, three dimensions is converted into two-dimensional space: can derive the colourity of pixel and linear relationship according to the relation of diffuse reflection and direct reflection against intensity:
P=m wherein
d(Λ-Г), σ is the colourity of pixel, so at 1/ ∑ I
cOpen in the contrary Strength Space with σ, all pixels constitute straight line, and intercept is exactly the color Г of light source;
Step D. adopts the parameter approximatioss that the point in the contrary Strength Space is carried out fitting a straight line, tries to achieve the chrominance axis intercept, promptly is the light source colour value; In contrary Strength Space, y=kx+b, x are 1/ ∑ I
c, y is σ, and k is a slope, and intercept b promptly is light source colour Г
c, MN is the perpendicular bisector of bounding box, the intersection point of MN and best-fitting straight line is E, for any 1 P on the straight line MN, be k straight line l through P point slope, all data points arrive the distance of straight line l and are d, must there be k=a so, a ∈ (∞ ,+∞), make the d value minimum, light source colour is certainly for just, that is to say intercept certainly greater than 0, so the slope span can be reduced into (0, y
Max/ x
Min), for arbitrary fixed slope k, go up the straight line l of point through MN
*, all data points are to straight line l
*Distance and be d
*, must have y=b so, b ∈ (y
Min, y
Max), make d
*Value is minimum, and so, the fitting a straight line problem just is converted to parameter k, the approximation problem of y, wherein y
Min, y
MaxThe minimum value of all pixel colourities and maximum, x
MaxMinimum value for the contrary intensity of all pixels;
Step e. according to the light source colour value, on normalized image, carry out color correction, recover image then: according to the light source colour that calculates, carry out color correction, just obtained the image behind the Light source correction, i.e. image under the standard white light.
Beneficial effect: compared with prior art, the present invention utilizes the voting mechanism rapid extraction to go out highlight area, the sum of errors shortcoming consuming time that one side has avoided image segmentation to bring, can extract the highlight area on a plurality of objects on the other hand simultaneously, be used for the highlight area extraction that complex scene comprises complex texture and owes the texture object.
In addition, the present invention to contrary Strength Space, is converted into two dimension with three-dimensional computations with normalized high light pixel spot projection, utilizes the employing parameter to approach and carries out fitting a straight line, tries to achieve light source colour fast and accurately, greatly reduces computation complexity.
Compared with prior art, the present invention proposes a kind of image segmentation that do not need, to article surface vein without limits, fast accurate light source colour computational methods.Be very helpful for the accuracy that improves graphical analysis and identification, have very important significance for machine vision, industrial detection.
Description of drawings
Fig. 1 is the entire block diagram of light source calculating of the present invention and method for correcting image;
Fig. 2 is that parameter of the present invention is approached the fitting a straight line schematic diagram.
Embodiment
As shown in Figure 1, the input picture of light source calculating of the present invention and method for correcting image is by after the image normalization module, pixel after the normalization enters the high light detection module of ballot method, detected highlight area pixel back projection is to contrary Strength Space, utilize parameter to approach fitting a straight line to the point in the contrary Strength Space then and calculate light source colour, image is carried out color correction, outputting standard white light hypograph according to the light source colour that calculates then.
For ease of profound understanding technology contents of the present invention, the present invention is described in detail below in conjunction with accompanying drawing.
1. input color image is carried out normalization.The relation table of light source colour Г and color of object surface Λ is shown I (x)=m
d(x) Λ (x)+m
s(x) Г, the color value of each pixel among I (x) the presentation video I wherein, m
d(x) remarked pixel point x place object surfaces diffuse reflectance, Λ (x) remarked pixel point x place object surfaces color, m
s(x) remarked pixel point x place object surfaces specularity factor, Г represents light source colour, and the color of each pixel x in the image I is carried out normalization: σ (x)=I (x)/∑ I
c(x), wherein σ (x) is the colourity of pixel x after the normalization.∑ I
c(x) three passage color values of remarked pixel x sum.
2. for the surf zone of same color, the pixel of pure diffuse reflection or a small amount of direct reflection and the variation of surrounding pixel point are mild, and the similarity of pixel is very big, utilizes this character, and we carry out high light to all pixels and detect.Suppose (x-1, x x+1) are three adjacent pixels points, and they are in the diffuse reflection zone of same color, can derive that neighbor color distortion value is in the same color region:
Wherein (x-1 x) is the color distortion value of two neighbor x-1 and x to Λ.Note ε=| Λ (x, x+1)-Λ (x-1, x) |, determine by voting mechanism whether a pixel is the diffuse reflection point.We think when ε＜α so, for pixel x-1, x and x+1 for throwing a ticket respectively.Here α is the threshold value that is provided with in advance.After having inspected all pixels, poll is exactly the diffuse reflection point greater than the point of preset value λ.Poll is exactly a specular reflection point less than the point of preset value.Each pixel will be checked 6 times in this algorithm.Be considered to marginal point when the color of neighbor pixel and light intensity differ greatly in addition, also be excluded outside highlight area.
3. on the basis of normalized image, the highlight area pixel is projected contrary Strength Space.
Can derive the colourity of pixel and the linear relationship of contrary intensity according to the relation of diffuse reflection and direct reflection:
P=m wherein
d(Λ-Г), σ is the colourity of pixel.So at 1/ ∑ I
cOpen in the contrary Strength Space with σ, all pixels constitute straight line, and intercept is exactly the color Г of light source.
4. for having a few in the contrary Strength Space, adopt the parameter approximatioss to carry out fitting a straight line, try to achieve light source colour.In contrary Strength Space, for the side of description we in 3 formula be rewritten as y=kx+b, x is 1/ ∑ I
c, y is σ, and k is a slope, and intercept b promptly is light source colour Г
cABCD is the bounding box of all data points as shown in Figure 2, and MN is the perpendicular bisector of bounding box, and the intersection point of MN and best-fitting straight line is E.For any 1 P on the straight line MN, be k straight line l through P point slope, all data points arrive the distance of straight line l and are d, must have k=a so, a ∈ (∞ ,+∞), make the d value minimum.Light source colour that is to say intercept certainly greater than 0 certainly for just, so the slope span can be reduced into (0, y
Max/ x
Min).For arbitrary fixed slope k, go up the straight line l of point through MN
*, all data points are to straight line l
*Distance and be d
*, must have y=b so, b ∈ (y
Min, y
Max), make d
*Value is minimum.So, the fitting a straight line problem just is converted to parameter k, the approximation problem of y.Y wherein
Min, y
MaxThe minimum value of all pixel colourities and maximum, x
MaxMinimum value for the contrary intensity of all pixels.
5. according to the light source colour that calculates, carry out color correction, just obtained the image behind the Light source correction, i.e. image under the standard white light.
Claims (1)
1. a light source colour calculates and method for correcting image, it is characterized in that this method may further comprise the steps:
Steps A. the colourity to each pixel in the coloured image is carried out normalization: the relation table of light source colour Γ and color of object surface Λ is shown I (x)=m
d(x) Λ (x)+m
s(x) Γ, the color value of each the pixel x among I (x) the presentation video I wherein, m
d(x) remarked pixel point x place object surfaces diffuse reflectance, Λ (x) remarked pixel point x place object surfaces color, m
s(x) remarked pixel point x place object surfaces specularity factor, Γ represents light source colour, and the color of each pixel x in the image I is carried out normalization: σ (x)=I (x)/∑ I
c(x), wherein σ (x) is the colourity of pixel x after the normalization; ∑ I
c(x) three passage color values of remarked pixel x sum, I
cRemarked pixel point x is at the color value of c Color Channel;
Step B. utilizes the variation of the pixel of pure diffuse reflection or a small amount of direct reflection and surrounding pixel point mild, be the big character of similarity of pixel, go out highlight area by the voting method rapid extraction: for the surf zone of same color, the pixel of pure diffuse reflection or a small amount of direct reflection and the variation of surrounding pixel point are mild, the similarity of pixel is very big, all pixels is carried out high light detect adjacent three pixel (x-1 in the same color region, x, x+1) the pixel color difference value is:
Λ (x-1 wherein, x) be the color distortion value of two neighbor x-1 and x, note ε=| Λ (x, x+1)-Λ (x-1, x) |, determine by voting mechanism whether a pixel is the diffuse reflection point, think so when ε＜α, for pixel x-1, x and x+1 for throwing a ticket respectively, α is the threshold value that is provided with in advance here; After having inspected all pixels, poll is exactly the diffuse reflection point greater than the point of preset value λ, and poll is exactly a specular reflection point less than the point of preset value; Each pixel will be checked 6 times in this algorithm, is considered to marginal point when the color of neighbor pixel and light intensity differ greatly in addition, also is excluded outside highlight area;
The bidirectional reflectance function illumination model of step C. from simplifying, derive contrary intensity and the linear relationship of colourity and the expression and the bearing calibration of light source colour, the chromatic value of three passages of the pixel in the highlight area is projected respectively by in contrary intensity and the space that colourity is opened, three dimensions is converted into two-dimensional space: can derive the colourity of pixel and linear relationship according to the relation of diffuse reflection and direct reflection against intensity:
P=m wherein
d(Λ-Γ), σ is the colourity of pixel, so at 1/ ∑ I
cOpen in the contrary Strength Space with σ, all pixels constitute straight line, and intercept is exactly the color Γ of light source;
Step D. adopts the parameter approximatioss that the point in the contrary Strength Space is carried out fitting a straight line, tries to achieve the chrominance axis intercept, promptly is the light source colour value; In contrary Strength Space, y=kx+b, x are 1/ ∑ I
c, y is σ, and k is a slope, and intercept b promptly is light source colour Γ
c, MN is the perpendicular bisector of bounding box, the intersection point of MN and best-fitting straight line is E, for any 1 P on the straight line MN, be k straight line l through P point slope, all data points arrive the distance of straight line l and are d, must there be k=a so, a ∈ (∞ ,+∞), make the d value minimum, light source colour is certainly for just, that is to say intercept certainly greater than 0, so the slope span can be reduced into (0, y
Max/ x
Min), for arbitrary fixed slope k, go up the straight line l of point through MN
*, all data points are to straight line l
*Distance and be d
*, must have y=b so, b ∈ (y
Min, y
Max), make d
*Value is minimum, and so, the fitting a straight line problem just is converted to parameter k, the approximation problem of y, wherein y
Min, y
MaxThe minimum value of all pixel colourities and maximum, x
MaxMinimum value for the contrary intensity of all pixels;
Step e. according to the light source colour value, on normalized image, carry out color correction, recover image then: according to the light source colour that calculates, carry out color correction, just obtained the image behind the Light source correction, i.e. image under the standard white light.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101332729A CN101146233A (en) | 2007-09-26 | 2007-09-26 | A computing and image correction method for light source color |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101332729A CN101146233A (en) | 2007-09-26 | 2007-09-26 | A computing and image correction method for light source color |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101146233A true CN101146233A (en) | 2008-03-19 |
Family
ID=39208480
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101332729A Pending CN101146233A (en) | 2007-09-26 | 2007-09-26 | A computing and image correction method for light source color |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101146233A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101635053B (en) * | 2009-09-02 | 2011-10-26 | 北京航空航天大学 | Method for calibrating color intensity consistency of multiple light sources |
| CN102301392A (en) * | 2009-01-29 | 2011-12-28 | 日本电气株式会社 | Color image processing method, color image processing device, and recording medium |
| CN102572211A (en) * | 2010-12-20 | 2012-07-11 | 三星泰科威株式会社 | Method and apparatus for estimating light source |
| CN102301391B (en) * | 2009-01-27 | 2013-07-17 | 日本电气株式会社 | Color image processing method, color image processing device and program |
| CN106408617A (en) * | 2016-08-26 | 2017-02-15 | 珠海金山网络游戏科技有限公司 | Interactive single image material acquiring system based on YUV color space and method |
| CN107251554A (en) * | 2015-02-09 | 2017-10-13 | （株）奈克斯其布斯 | Color table creating device, camera review correction/control device and its method |
| CN109672874A (en) * | 2018-10-24 | 2019-04-23 | 福州大学 | A kind of consistent three-dimensional video-frequency color calibration method of space-time |
| WO2019085711A1 (en) * | 2017-10-30 | 2019-05-09 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | White balance processing method, electronic device and computer readable storage medium |
| CN109829480A (en) * | 2019-01-04 | 2019-05-31 | 广西大学 | The method and system of the detection of body surface bloom feature and material classification |
| CN110060308A (en) * | 2019-03-28 | 2019-07-26 | 杭州电子科技大学 | A kind of color constancy method based on light source colour distribution limitation |
-
2007
- 2007-09-26 CN CNA2007101332729A patent/CN101146233A/en active Pending
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102301391B (en) * | 2009-01-27 | 2013-07-17 | 日本电气株式会社 | Color image processing method, color image processing device and program |
| CN102301392A (en) * | 2009-01-29 | 2011-12-28 | 日本电气株式会社 | Color image processing method, color image processing device, and recording medium |
| US8693773B2 (en) | 2009-01-29 | 2014-04-08 | Nec Corporation | Color image processing method, color image processing device, and recording medium |
| CN102301392B (en) * | 2009-01-29 | 2014-08-13 | 日本电气株式会社 | Color image processing method, color image processing device, and recording medium |
| CN101635053B (en) * | 2009-09-02 | 2011-10-26 | 北京航空航天大学 | Method for calibrating color intensity consistency of multiple light sources |
| CN102572211A (en) * | 2010-12-20 | 2012-07-11 | 三星泰科威株式会社 | Method and apparatus for estimating light source |
| CN102572211B (en) * | 2010-12-20 | 2015-11-25 | 韩华泰科株式会社 | Estimate the method and apparatus of light source |
| CN107251554A (en) * | 2015-02-09 | 2017-10-13 | （株）奈克斯其布斯 | Color table creating device, camera review correction/control device and its method |
| CN106408617A (en) * | 2016-08-26 | 2017-02-15 | 珠海金山网络游戏科技有限公司 | Interactive single image material acquiring system based on YUV color space and method |
| CN106408617B (en) * | 2016-08-26 | 2020-03-31 | 珠海金山网络游戏科技有限公司 | Interactive single image material obtaining system and method based on YUV color space |
| WO2019085711A1 (en) * | 2017-10-30 | 2019-05-09 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | White balance processing method, electronic device and computer readable storage medium |
| US10616544B2 (en) | 2017-10-30 | 2020-04-07 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | White balance processing method, electronic device and computer readable storage medium |
| US10674128B2 (en) | 2017-10-30 | 2020-06-02 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | White balance processing method, electronic device and computer readable storage medium |
| CN109672874A (en) * | 2018-10-24 | 2019-04-23 | 福州大学 | A kind of consistent three-dimensional video-frequency color calibration method of space-time |
| CN109672874B (en) * | 2018-10-24 | 2020-08-11 | 福州大学 | Space-time consistent stereo video color correction method |
| CN109829480A (en) * | 2019-01-04 | 2019-05-31 | 广西大学 | The method and system of the detection of body surface bloom feature and material classification |
| CN110060308A (en) * | 2019-03-28 | 2019-07-26 | 杭州电子科技大学 | A kind of color constancy method based on light source colour distribution limitation |
| CN110060308B (en) * | 2019-03-28 | 2021-02-02 | 杭州电子科技大学 | Color constancy method based on light source color distribution limitation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101146233A (en) | A computing and image correction method for light source color | |
| Suo et al. | Fast and high quality highlight removal from a single image | |
| Artusi et al. | A survey of specularity removal methods | |
| Klinker et al. | A physical approach to color image understanding | |
| CN100555309C (en) | Be used for determining the method for direction of the main light source of image | |
| EP1143375B1 (en) | Device, method and record medium for image comparison | |
| US9639748B2 (en) | Method for detecting persons using 1D depths and 2D texture | |
| CN112052831B (en) | Method, device and computer storage medium for face detection | |
| CN104156693B (en) | A kind of action identification method based on the fusion of multi-modal sequence | |
| Fang et al. | A method to segment moving vehicle cast shadow based on wavelet transform | |
| Zhou et al. | Radiometric invariant stereo matching based on relative gradients | |
| Lopez et al. | A new preprocessing technique for fast hyperspectral endmember extraction | |
| Tian et al. | Human Detection using HOG Features of Head and Shoulder Based on Depth Map. | |
| Cruz et al. | Aerial detection in maritime scenarios using convolutional neural networks | |
| CN106980371B (en) | It is a kind of based on the mobile augmented reality exchange method for closing on heterogeneous distributed structure | |
| Antensteiner et al. | Single image multi-spectral photometric stereo using a split u-shaped cnn | |
| Wesolkowski et al. | Shading-and highlight-invariant color image segmentation using the MPC algorithm | |
| Youseph et al. | Pixel and edge based illuminant color estimation for image forgery detection | |
| Ettalibi et al. | AI and Computer Vision-based Real-time Quality Control: A Review of Industrial Applications | |
| KR20200082074A (en) | Method for acquiring illumination invariant image using color space compression technique and device therefor | |
| WO2022272166A1 (en) | Systems and methods for generating and using visual datasets for training computer vision models | |
| Li et al. | Spatio-context-based target tracking with adaptive multi-feature fusion for real-world hazy scenes | |
| Mazumdar et al. | Forgery detection in digital images through lighting environment inconsistencies | |
| EP3493155A1 (en) | Surface color segmentation | |
| Ling et al. | Infrared target tracking with kernel-based performance metric and eigenvalue-based similarity measure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Open date: 20080319 |