CN106895916A - A kind of single exposure shoots the method for obtaining multispectral image - Google Patents
A kind of single exposure shoots the method for obtaining multispectral image Download PDFInfo
- Publication number
- CN106895916A CN106895916A CN201710013391.4A CN201710013391A CN106895916A CN 106895916 A CN106895916 A CN 106895916A CN 201710013391 A CN201710013391 A CN 201710013391A CN 106895916 A CN106895916 A CN 106895916A
- Authority
- CN
- China
- Prior art keywords
- matrix
- trn
- rgb
- target
- measured
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 47
- 238000002310 reflectometry Methods 0.000 claims abstract description 43
- 238000001228 spectrum Methods 0.000 claims abstract description 34
- 238000003384 imaging method Methods 0.000 claims abstract description 6
- 239000011159 matrix material Substances 0.000 claims description 62
- 230000003595 spectral effect Effects 0.000 claims description 40
- 238000005457 optimization Methods 0.000 claims description 12
- 230000008859 change Effects 0.000 claims description 3
- 238000000205 computational method Methods 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 102220326096 rs1378670216 Human genes 0.000 claims description 2
- 102220015909 rs138331646 Human genes 0.000 claims description 2
- 102220304695 rs944054671 Human genes 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 13
- 230000000694 effects Effects 0.000 abstract description 4
- 230000003287 optical effect Effects 0.000 abstract description 4
- 238000012216 screening Methods 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 abstract description 2
- 238000002059 diagnostic imaging Methods 0.000 abstract description 2
- 239000000463 material Substances 0.000 abstract description 2
- 238000007639 printing Methods 0.000 abstract description 2
- 230000003796 beauty Effects 0.000 abstract 1
- 238000001914 filtration Methods 0.000 abstract 1
- 230000004075 alteration Effects 0.000 description 11
- 238000012360 testing method Methods 0.000 description 7
- 238000005259 measurement Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000000701 chemical imaging Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 230000004907 flux Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241000227425 Pieris rapae crucivora Species 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000000985 reflectance spectrum Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000010563 solid-state fermentation Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J3/2823—Imaging spectrometer
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J3/2823—Imaging spectrometer
- G01J2003/2826—Multispectral imaging, e.g. filter imaging
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
- G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
- G01J3/28—Investigating the spectrum
- G01J2003/283—Investigating the spectrum computer-interfaced
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- General Physics & Mathematics (AREA)
- Spectrometry And Color Measurement (AREA)
Abstract
The method for obtaining multispectral image is shot the invention discloses a kind of single exposure, belongs to optical engineering field.Algorithm proposed by the present invention can utilize multinomial model, adaptively selected training sample, automatic different weights can also be assigned to different training samples, and the spectrum of lighting source can also be optimized to obtain optimal reflectivity reconstruction effect, realize the single exposure by common RGB camera under a kind of light source to shoot, it becomes possible to get multispectral image.The present invention has quicker, it is not necessary to filtering device, reduces mechanical action, the advantage of similar precision, and it is capable of the spectrum of capturing video and the spectrum of dynamic object, can uses in numerous applications, including but not limited to computer graphics or medical imaging, printing industry, historical relic reappears, bio-imaging, beauty industry, the fields such as material screening, and other spectrum field of reproduction such as colour reproduction.
Description
Technical field
The invention belongs to optical engineering field, and in particular to a kind of single exposure shoots the method for obtaining multispectral image.
Background technology
Spectral reflectance recovery is a key areas in optical research, is obtained the purpose is to pass through various imaging devices
The device-dependent RGB tristimulus values for taking reconstruct the object spectral reflectivity all unrelated with equipment and illumination inherently,
Can be used for Accurate Prediction object color under different lighting conditions.The spectral reflectivity of object, i.e., object is to different wave length
The ratio between the luminous flux of light reflection and the luminous flux of incidence.Multispectral image refers to a width picture, and each of which pixel is all
It is made up of spectral reflectivity.Multispectral image acquisition modes, at this stage for be divided into two kinds：First, object spectra reflectivity is obtained
Conventional method be to utilize spectrophotometer.But spectrophotometer once can only be very big to a measurement, its workload,
Then occur in that to efficiently measure the spectral reflectivity of every on image, it is possible to use with more than ten to tens passages
Multi-optical spectrum imaging system (Multispectral imaging systems, MSIS).The system is generally by charge-coupled device
Part, optical filter, rectilinear scanner, or the method based on multi-channel illumination is developed.The multispectral imaging of these types
System wastes time and energy, and expensive.As Application No. 201310222379.6,201310487513.5,
Device disclosed in 201180013153.4 patent, it, can be using interference filter, platform-type when multispectral image is obtained
Scanner or sensor, produce a series of monochromatic light of wavelength, the multispectral image being finally synthesizing.2nd, it is (logical based on multiexposure, multiple exposure
Often 3 times) RGB triple channel cameras, it generally produces suitable light source using different narrow-band LEDs, and by one kind
Computational methods, such as principal component analysis (Principle Component Analysis, PCA), wiener estimate (Weiner
Method), method of pseudo inverse matrix etc. rebuilds spectral reflectivity.Patent of invention such as Application No. 201310210413.8 is public
Opened a kind of multispectral image acquisition method, its need using colour filter/without colour filter time-division safety pin to high, medium and low exposure
Each shooting of amount is once；The response of the digital camera obtained after obtaining shooting for three times, then the fusion for carrying out high dynamic range images
Treatment；The data fusion of three shootings is obtained into the Image Coding corresponding with actual photographed object brightness, then by Image Coding
Data complete the reconstruction of spectral reflectivity.This method is than the calculating cost needed for the method for traditional utilization polylith filter more
It is low, but due to needing multiexposure, multiple exposure, be not suitable for shooting dynamic object or filmed image.So far, also not any
The method for finding to obtain multispectral image based on single exposure in patent or technical literature.
The content of the invention
Need to rely on complicated imaging system or many present invention aim to address multispectral image is obtained in the prior art
The brought inconvenience of secondary exposure, and a kind of method that single exposure shoots acquisition multispectral image is provided.It is of the present invention
Concrete technical scheme is as follows：
Single exposure shoots the method for obtaining multispectral image, and step is as follows：
S1：Selection is similar to the spectral quality of target to be measured (can be test sample, target object or target scene)
M training sample, determine its spectral reflectivity under K kind different wave lengths respectively, and as matrix column element, build
Into a matrix R of K × Mtrn, each of which row represent a K spectral reflectance values of training sample respectively.Training sample
Ensure the similitude with target to be measured, such as during training sample selection, it is considered to the representativeness and correlation of sample, i.e.,
Seeking training sample can represent whole sample space, while there is certain correlation with reconstruction sample, can preferably reflect
The spectral characteristic of target to be measured.Training sample can use standard color card, for exampleThe Macbeth of company
ColorChecker, Digital SG ColorChecker, or other any color libraries specified, it is also possible to rule of thumb
Or existing screening technique is selected.And the number M of training sample can be selected according to actual conditions, but need to meet
Sample size reaches the minimum requirements of training enough.Due to the spectral reflectivity curve tool of most of body surfaces in human lives
There is smooth property, so spectral reflectivity curve rate can take K in the range of the span lengths of spectrum and adopt in real process
Sampling point, whole piece spectral reflectivity curve rate is fitted with the spectral reflectance values at different wave length.The value of K can be according to actually entering
Row selection.
S2：Under lighting source, the rgb value that M training sample is imaged in a certain RGB camera is obtained, build 3 × M's
Training sample Matrix Ctrn, each of which row represent R, G and B value of training sample respectively.
S3：Respectively obtain N (N >=1) individual target to be measured under with S2 identical lighting sources, with S2 identical RGB phases
The rgb value being imaged in machine, is built into a Matrix C of 3 × Ntst, each of which row represent respectively R, the G of target to be measured with
And B values.
S4：Respectively to described Matrix CtrnAnd CtstIn R, G, B component carry out identical Polynomial Expansion, obtain P × M
Training sample extended matrix C'trnWith the extended matrix C' of P × Ntst.Polynomial Expansion refers to utilize polynomial module in the present invention
Type further expands its high order component, such as RG, RB, GB, R of second order based on the R in script matrix, G, B component2、G2、B2
With the R of three ranks2G, RGB etc., and 1.When carrying out Polynomial Expansion to matrix, its newly-increased vector value can be according to extension form
Carry out mathematical computations, such as R, G, B component on the basis of, extend R2, RGB and 1, then need to CtrnAnd CtstMatrix increases
Three rows, respectively as R2, RGB and 1 component, R2The value of a certain row of being expert at is exactly the flat of line number value where R component in the row
Side, RGB in being expert at the value of a certain row be exactly R in the row, G, the product of B component place line number value, the vector that 1 component is expert at
It is worth all 1.Other kinds of Polynomial Expansion calculation is similar.
S5：Based on the matrix obtained in abovementioned steps, the reflectivity of target to be measured is rebuild, obtain the multispectral of target to be measured
Image.The reflectivity method for reconstructing that can be used include but is not limited to it is following several, for example：
1) pseudo-inverse matrix method-pseudo-inverse method (Babaei, V., S.H.Amirshahi and
F.Agahian,Using weighted pseudo-inverse method for reconstruction of
reflectance spectra and analyzing the dataset in terms of normality.Color
Research&Application,2011.36(4):p.295–305)；2) the smoothness constraint method-curve based on curve smoothing
smoothing based smoothness constraint method(Li,C.and M.R.Luo.The Estimation
of Spectral Reflectances Using the Smoothness Constraint Condition.2001)；3) it is main
Componential analysis-principle component analysis based method (Sobagaki, H., K.Takahama
and Y.Nayatani,Estimation of spectral reflectance functions for Munsell
renotations.Journal of the Color Science Association of Japan,1989.13(2):
p.150-152)；4) the wiener estimation technique-Weiner ' s estimation based method (Stigell, P., K.Miyata
and M.Hauta-Kasari,Wiener estimation method in estimating of spectral
reflectance from RGB images.Pattern Recognition and Image Analysis,2007.17
(2):p.233-242)。
As a kind of preferred embodiment, can be adopted with the following method during the multispectral image for calculating target to be measured:
Rrec=Rtrn(C'trn)⊥C'tst
Wherein RrecIt is the reflection rate matrix rebuild, each row represent a target to be measured in the K light of different wave length respectively
Spectrum reflectivity；Symbol ⊥ represents that this is a pseudo inverse matrix.
As another preferred embodiment, can also be to Matrix CtrnAnd RtrnColumn vector carries out identical extension, by former square
The i-th row in battle array repeat wiTime, wherein W=w1+w2+……+wMBe the total degree that all samples repeat, finally give P ×
The training sample extended matrix C' of WtrnWith the extended matrix R' of K × Wtrn.With RtrnAs a example by, the i-th row in original matrix are repeated into wi
After, may be characterized asWhereinIt is consistent with the i-th column vector in original matrix.In optimization process, one should be kept to two extensions of matrix
Cause, i.e., the pass w that the i-th row are expanded in the matrix after two extensionsiShould be consistent.It is expanded after matrix, with the square after extension
Battle array replaces original matrix to calculate the multispectral image of target to be measured, and computational methods are Rrec=R'trn(C'trn)⊥C'tst.Then, constantly
To wiOptimization is iterated, until obtaining optimal multispectral image.During iteration optimization, reflectivity and reality that can be to rebuild
Aberration or mean square error between reflectivity are optimizing index, it would however also be possible to employ other optimizing index, are not limited herein.
Further, present invention also offers a kind of spectrum optimized algorithm of self adaptation, obtained using the optimized algorithm
Spectrum, can obtain the precision higher than other light sources.The step of algorithm is optimized to lighting source is as follows：
An any given initial lighting source, the reflectivity of target to be measured is rebuild according to step S1~S5；Then not
The disconnected spectral composition changed in light source, finally gives the lighting source with optimal spectrum.The used optimizing index of optimization with
It is foregoing the same, can be selected according to actual conditions.In addition, it is necessary to, it is noted that the optimization of lighting source can be carried out individually,
Can combine with the extended mode of foregoing column vector is carried out.It is further preferred that constantly to wiOptimization is iterated, until obtaining
After optimization after multispectral image, then using spectrum optimized algorithm, spectral composition is optimized, obtain optimal multispectral figure
Picture.Certainly, both sequentially may be reversed, or synchronously carry out.Optimized algorithm includes but is not limited to particle cluster algorithm, and heredity is calculated
Method etc..
As another preferred embodiment, in described Polynomial Expansion, with the change of P values using different multinomial
Formula；If multinomial during note=3 is P3=[R G B], during P=5, Polynomial Expansion is P5=[P3 RGB 1]；During P=9
Polynomial Expansion is P9=[P3 RG RB GB R2 G2 B2], Polynomial Expansion is P11=[P9 RGB 1], P=during P=11
Widenable to P18=[P11 RG when 182 RB2 GR2 GB2 BR2 BG2 R3 G3 B3], if widenable to P20=during P=20
[P18 RGB 1].Preferably it is to make P=11, Matrix CtrnAnd CtstIn R, G, B component by Polynomial Expansion be R, G, B, R2,
G2,B2, RG, GB, BR, RGB and 1, the multinomial model provides sufficient color component, preferably can be arrived for RGB
The mapping of XYZ, improves precision.
As another preferred embodiment, in described S2, obtain what M training sample was imaged in a certain RGB camera
The method of rgb value can use following two modes：One is under with S1 identical lighting sources, by the RGB phases in S2
Machine directly shoots the image of training sample, and its rgb value is obtained according to the image for obtaining；The second is using the RGB camera in S2
The spectral power distribution of lighting source in spectrum sensitive function (spectral sensitivity functions, SSFs), S1
The spectral reflectivity of function and training sample, calculates the rgb value that training sample is imaged in RGB camera.The spectrum sensitive of camera
Function can be obtained by manufacturer, it is also possible to be determined by testing.
In general, lighting source meets following condition：There is certain spectrum energy in K wavelength of albedo measurement
Amount.Common white light source usually can meet the requirement, therefore need not be using complicated, expensive special light sources.Certainly, it is actual
On, if light source only has monochromatic light, then also can correctly reduce the reflectance value of the wave band, but the precision of its all band can be very low
The present invention is realized the single exposure by common RGB camera under a kind of light source and is shot by the algorithm for optimizing,
Multispectral image can be just got, so as to greatly reduce in multispectral image acquisition process to equipment, the requirement of light source.This
Invention can instead of three kinds of LED of red, green, blue using single lighting source (such as white light source).Algorithm proposed by the present invention
Multinomial model is used, adaptively selected training sample automatic can also assign different weights to different training samples,
And the spectrum of lighting source can also be optimized to obtain optimal reflectivity reconstruction effect.The present invention have more accelerate
Speed, low cost, the advantage of similar precision, and it is capable of the spectrum of capturing video and the spectrum of dynamic object, can be used in
In various applications, including but not limited to computer graphics or medical imaging, printing industry, historical relic reproduction, bio-imaging are beautiful
Hold the fields such as industry, material screening, and other spectrum field of reproduction such as colour reproduction.
Specific effect of the invention will be described in detail by follow-up embodiment.
Brief description of the drawings
Fig. 1 be the embodiment of the present invention in obtain multispectral image system structure diagram；
Fig. 2 is the 4000K light sources used in embodiment；
Fig. 3 is the optimal spectrum obtained in embodiment 4.
Specific embodiment
Patent of the present invention is further described below in conjunction with the accompanying drawings, because facilitating a better understanding of.In patent of the present invention
Technical characteristic on the premise of not colliding with each other, can be mutually combined, be not construed as limiting.
As shown in figure 1, to be used for the system for obtaining multispectral image in following embodiments, system composition includes one
The digital camera 1 (Canon's 5D2 cameras, to photographed scene 4) of RGB triple channels, a lighting apparatus 2 (is simulated using LED
4000K lighting source, the LED that the light source can be constituted by changing adjusts spectral composition), spectral measurement device,
Such as spectroradiometer (being used to measure the spectral power distribution of lighting source) a, control computer 3 (is used to control device
And calculate), and for the training sample of optimized algorithm performance.In each embodiment, the spectrum of lighting source is by Radiation intensity
Meter (model：The UV tele-spectroradiometer of JETI Specbos 1211) measure.Train the spectrum of sample anti-
Rate is penetrated to be measured by spectrophotometer (Datacolor SF600).It is pointed out that the system is only a kind of realization side
Formula, those skilled in the art can also be variously changed and optimize, the method being not intended to limit the present invention.
Embodiment 1
Conventional method is used in the present embodiment, multinomial is neither carried out and is expanded nor carry out weight optimization, exposed by single
Light shoots and obtains multispectral image.Comprise the following steps that：
S1：The test sample and training sample for being used are all from Digital SG ColorChecker, wherein colored color
Sample amounts to 96.It is classified as two parts in experiment, test sample and each 48 of sample of training.From in 400-700nm wave-length coverages
With 400nm as starting point, every 10nm, a sampled point, altogether 31 are set.Determine 48 training respectively by spectrophotometer
Spectral reflectivity of the sample under 31 sampled points, and as matrix row element, it is built into the matrix of 31 × 48
Rtrn, each of which row represent the 31 of training sample spectral reflectance values respectively；
S2：The spectral composition of lighting source is as shown in Fig. 2 under this light source, using the spectrum sensitive function of camera, obtain
48 rgb values of training sample, build 3 × 48 training sample Matrix Ctrn, each of which row represent a training sample respectively
R, G and B value；
S3：Under with S2 identical lighting sources, by with S2 identical RGB cameras, in digital camera be imaged, obtain
The rgb value of test sample, is built into the Matrix C of 3 × 48tst, each of which row represent R, the G of target to be measured respectively
And B values；
S4：The reflectivity of target to be measured is rebuild, the multispectral image of target to be measured is obtained：Calculate the multispectral of target to be measured
The method of image is：
Rrec=Rtrn(Ctrn)⊥Ctst
Wherein RrecIt is the reflection rate matrix rebuild, each row represent a target to be measured in the K light of different wave length respectively
Spectrum reflectivity；Symbol ⊥ represents that this is a pseudo inverse matrix.
Aberration is calculated using CIEDE2000 colour difference formulas, the reflectivity of the reflectivity and reality rebuild is calculated in D65 standards
Light source, standard A light sources, and the value of chromatism under F11 light sources, as a result for：Average color difference 5.5, maximum aberration 13, standard deviation 2.2.
Embodiment 2
Only with Polynomial Expansion in the present embodiment, shot by single exposure and obtain multispectral image.Specific steps are such as
Under：
S1：This step is same as Example 1.
S2：The spectral composition of lighting source is as shown in Fig. 2 under this light source, using the spectrum sensitive function of camera, obtain
48 rgb values of training sample, build 3 × 48 training sample Matrix Ctrn, each of which row represent a training sample respectively
R, G and B value；
S3：Under with S2 identical lighting sources, by with S2 identical RGB cameras, in digital camera be imaged, obtain
The rgb value of test sample, is built into the Matrix C of 3 × 48tst, each of which row represent R, the G of target to be measured respectively
And B values；
S4：Respectively to described Matrix CtrnAnd CtstIn R, G, B component carry out identical Polynomial Expansion, wherein P=
11, R, G, B component are by including R, G, B, R after extension2,G2,B2, RG, GB, BR, RGB and 1 obtain 11 × 48 training sample
This extended matrix C'trnExtended matrix C' with 11 × 48tst；
S5：The reflectivity of target to be measured is rebuild, the multispectral image of target to be measured is obtained.Calculate the multispectral of target to be measured
The method of image is:
Rrec=Rtrn(C'trn)⊥C'tst
Wherein RrecIt is the reflection rate matrix rebuild, each row represent a target to be measured in the K light of different wave length respectively
Spectrum reflectivity；Symbol ⊥ represents that this is a pseudo inverse matrix.
Aberration is calculated using CIEDE2000 colour difference formulas, the reflectivity of the reflectivity and reality rebuild is calculated in D65 standards
Light source, standard A light sources, and the value of chromatism under F11 light sources, as a result for：Average color difference 3.2, maximum aberration 10.8, standard deviation
1.2。
Embodiment 3
In the present embodiment, using Polynomial Expansion is first carried out, weight optimization mode is then carried out.Comprise the following steps that：
S1：This step is same as Example 1.
S2：The spectral composition of lighting source is as shown in Fig. 2 under this light source, using the spectrum sensitive function of camera, obtain
48 rgb values of training sample, build 3 × 48 training sample Matrix Ctrn, each of which row represent a training sample respectively
R, G and B value；
S3：Under with S2 identical lighting sources, by with S2 identical RGB cameras, in digital camera be imaged, obtain
The rgb value of test sample, is built into the Matrix C of 3 × 48tst, each of which row represent R, the G of target to be measured respectively
And B values；
S4：Respectively to described Matrix CtrnAnd CtstIn R, G, B component carry out identical Polynomial Expansion, after extension point
Amount number is R, G, B, R2,G2,B2, RG, GB, BR, RGB and 1 obtain 11 × 48 training sample extended matrix C "trnWith 11 ×
48 extended matrix C'tst；
Then, to Matrix C "trnAnd RtrnColumn vector carries out identical extension, by the initial repetition of the i-th row in original matrix
wiTime, obtain the training sample extended matrix C' of 11 × WtrnWith the extended matrix R' of 31 × Wtrn, W is wiSummation.
S5：The reflectivity of target to be measured is rebuild, the multispectral image of target to be measured is obtained, the multispectral of target to be measured is calculated
The method of image is:
Rrec=Rtrn(C'trn)⊥C'tst
Wherein RrecIt is the reflection rate matrix rebuild, each row represent a target to be measured in the K light of different wave length respectively
Spectrum reflectivity；Symbol ⊥ represents that this is a pseudo inverse matrix.
Then, it is optimizing index with the aberration between the reflectivity and the reflectivity of reality rebuild, constantly to the value of each wi
Optimization is iterated, until obtaining optimal multispectral image.
Aberration is calculated using CIEDE2000 colour difference formulas, the reflectivity of the reflectivity and reality rebuild is calculated in D65 standards
Light source, standard A light sources, and the value of chromatism under F11 light sources, as a result for：Average color difference 1.9, maximum aberration 7.5, standard deviation
1.4。
Embodiment 4
In the present embodiment, based on embodiment 3, continue using the spectrum optimized algorithm of self adaptation, to the initial photograph shown in Fig. 2
Spectral composition in Mingguang City source is optimized.Constantly change the spectral composition in light source, repeat step S1~S5, using population
Algorithm (or genetic algorithm) optimizes spectral composition, until obtaining.The spectral composition of the lighting source for finally giving is as shown in Figure 3.
By the optimal spectrum of use in embodiment 3, it is consistent including remaining parameter including the wi in weight optimization, obtains it
Calculate the aberration of the reflectivity and reflectivity of reality rebuild.
Aberration is calculated using CIEDE2000 colour difference formulas, calculates the reflectivity of the reflectivity and reality rebuild in D65 standards
Light source, standard A light sources, and the value of chromatism under F11 light sources, as a result for：Average color difference 0.4, maximum aberration 1.7, standard deviation
0.3。
Its specific effect is as shown in the table：
Embodiment described above is a kind of preferably scheme of patent of the present invention, and so it is not intended to limiting the invention
Patent.About the those of ordinary skill of technical field, in the case where the spirit and scope of patent of the present invention are not departed from, can be with
Make a variety of changes and modification.Therefore the technical scheme that all modes for taking equivalent or equivalent transformation are obtained, all falls within
In the protection domain of patent of the present invention.
Claims (6)
1. a kind of single exposure shoots the method for obtaining multispectral image, it is characterised in that step is as follows：
S1：M training sample for selecting the spectral quality of target to be measured similar, determines its light under K different wave length respectively
Spectrum reflectivity, and as matrix column element, it is built into a matrix R of K × Mtrn, each of which row represent one respectively
K spectral reflectance values of training sample；
S2：Under lighting source, the rgb value that M training sample is imaged in a certain RGB camera is obtained, build the training of 3 × M
Sample matrix Ctrn, each of which row represent R, G and B value of training sample respectively；
S3：N number of target to be measured is obtained respectively under with S2 identical lighting sources, is imaged in S2 identical RGB cameras
Rgb value, is built into a Matrix C of 3 × Ntst, each of which row represent R, G and B value of target to be measured respectively；
S4：Respectively to described Matrix CtrnAnd CtstIn R, G, B component carry out identical Polynomial Expansion, obtain the instruction of P × M
Practice sample extended matrix C'trnWith the extended matrix C' of P × Ntst；
S5：The reflectivity of target to be measured is rebuild, the multispectral image of target to be measured is obtained.
2. the method for claim 1, it is characterised in that the method for calculating the multispectral image of target to be measured is:
Rrec=Rtrn(C'trn)⊥C'tst
Wherein RrecIt is the reflection rate matrix rebuild, it is anti-in the spectrum of K different wave length that each row represent a target to be measured respectively
Penetrate rate；Symbol ⊥ represents that this is a pseudo inverse matrix.
3. the method for claim 1, it is characterised in that to Matrix CtrnAnd RtrnColumn vector carries out identical extension,
The i-th row in original matrix are repeated into wiTime, finally give the training sample extended matrix C' of P × WtrnWith the extended matrix of K × W
R'trn, then the multispectral image of target to be measured is calculated, computational methods are Rrec=R'trn(C'trn)⊥C'tst；Then to wiChanged
Generation optimization, until obtaining optimal multispectral image.
4. the method as described in claim 1 or 3, it is characterised in that optimized to described lighting source, step is as follows：
An any given initial lighting source, the reflectivity of target to be measured is rebuild according to step S1~S5；Then continuous iteration changes
Spectral composition in changing light, finally gives the lighting source with optimal spectrum.
5. the method for claim 1, it is characterised in that in described Polynomial Expansion, used with the change of P values
Different multinomials；If multinomial during note=3 is P3=[R G B]；If during P=5, Polynomial Expansion is P5=[P3 RGB
1]；If Polynomial Expansion is P9=[P3 RG RB GB R during P=92 G2 B2]；If Polynomial Expansion is P11=during P=11
[P9 RGB 1]；If widenable to P18=[P11 RG during P=182 RB2 GR2 GB2 BR2 BG2 R3 G3 B3]；If P=20
When widenable to P20=[P18 RGB 1].
6. the method for claim 1, it is characterised in that in described S2, obtains M training sample in a certain RGB
The method of the rgb value of imaging is in camera：Under with S1 identical lighting sources, instruction is directly shot by the RGB camera in S2
Practice the image of sample, its rgb value is obtained according to image；Or using the illumination in spectrum sensitive function, the S1 of the RGB camera in S2
The function of spectral power distribution of light source and the spectral reflectivity of training sample, calculate the RGB that training sample is imaged in RGB camera
Value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710013391.4A CN106895916B (en) | 2017-01-09 | 2017-01-09 | A kind of method that single exposure shooting obtains multispectral image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710013391.4A CN106895916B (en) | 2017-01-09 | 2017-01-09 | A kind of method that single exposure shooting obtains multispectral image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106895916A true CN106895916A (en) | 2017-06-27 |
| CN106895916B CN106895916B (en) | 2018-10-30 |
Family
ID=59198467
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710013391.4A Active CN106895916B (en) | 2017-01-09 | 2017-01-09 | A kind of method that single exposure shooting obtains multispectral image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106895916B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107505046A (en) * | 2017-07-31 | 2017-12-22 | 征图新视（江苏）科技有限公司 | Multispectral color measurement method |
| CN110399904A (en) * | 2019-06-28 | 2019-11-01 | 湖南工业大学 | A kind of emission spectrum method for reconstructing based on full-color fluorescence printing |
| CN110475072A (en) * | 2017-11-13 | 2019-11-19 | Oppo广东移动通信有限公司 | Shoot method, apparatus, terminal and the storage medium of image |
| WO2023273411A1 (en) * | 2021-07-01 | 2023-01-05 | 奥比中光科技集团股份有限公司 | Multispectral data acquisition method, apparatus and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103528968A (en) * | 2013-11-01 | 2014-01-22 | 上海理工大学 | Reflectance spectrum reconstruction method based on iterative threshold method |
| CN104634745A (en) * | 2015-01-29 | 2015-05-20 | 上海理工大学 | Spectral reconstruction algorithm |
| CN104634451A (en) * | 2015-02-11 | 2015-05-20 | 武汉大学 | Spectrum reconstruction method and system based on multichannel imaging system |
| CN105069234A (en) * | 2015-08-13 | 2015-11-18 | 武汉大学 | Spectrum dimensionality reduction method and system based on visual perception feature |
| CN106153193A (en) * | 2016-07-22 | 2016-11-23 | 浙江大学 | A kind of method utilizing the double light source response value of multispectral camera to obtain spectral reflectance |
-
2017
- 2017-01-09 CN CN201710013391.4A patent/CN106895916B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103528968A (en) * | 2013-11-01 | 2014-01-22 | 上海理工大学 | Reflectance spectrum reconstruction method based on iterative threshold method |
| CN104634745A (en) * | 2015-01-29 | 2015-05-20 | 上海理工大学 | Spectral reconstruction algorithm |
| CN104634451A (en) * | 2015-02-11 | 2015-05-20 | 武汉大学 | Spectrum reconstruction method and system based on multichannel imaging system |
| CN105069234A (en) * | 2015-08-13 | 2015-11-18 | 武汉大学 | Spectrum dimensionality reduction method and system based on visual perception feature |
| CN106153193A (en) * | 2016-07-22 | 2016-11-23 | 浙江大学 | A kind of method utilizing the double light source response value of multispectral camera to obtain spectral reflectance |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107505046A (en) * | 2017-07-31 | 2017-12-22 | 征图新视（江苏）科技有限公司 | Multispectral color measurement method |
| CN107505046B (en) * | 2017-07-31 | 2020-06-12 | 征图新视(江苏)科技股份有限公司 | Multispectral color measurement method |
| CN110475072A (en) * | 2017-11-13 | 2019-11-19 | Oppo广东移动通信有限公司 | Shoot method, apparatus, terminal and the storage medium of image |
| CN110475072B (en) * | 2017-11-13 | 2021-03-09 | Oppo广东移动通信有限公司 | Method, device, terminal and storage medium for shooting image |
| US11412153B2 (en) | 2017-11-13 | 2022-08-09 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Model-based method for capturing images, terminal, and storage medium |
| CN110399904A (en) * | 2019-06-28 | 2019-11-01 | 湖南工业大学 | A kind of emission spectrum method for reconstructing based on full-color fluorescence printing |
| WO2023273411A1 (en) * | 2021-07-01 | 2023-01-05 | 奥比中光科技集团股份有限公司 | Multispectral data acquisition method, apparatus and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106895916B (en) | 2018-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Oh et al. | Do it yourself hyperspectral imaging with everyday digital cameras | |
| Nguyen et al. | Training-based spectral reconstruction from a single RGB image | |
| JP5883949B2 (en) | Spectral synthesis for image capture device processing. | |
| CN106895916B (en) | A kind of method that single exposure shooting obtains multispectral image | |
| Shrestha et al. | One-shot multispectral color imaging with a stereo camera | |
| CN104168475B (en) | A kind of imaging type of digital camera changeable parameters obtains the method for color tristimulus values | |
| Shrestha et al. | Multispectral imaging using LED illumination and an RGB camera | |
| CN113676628B (en) | Image forming apparatus and image processing method | |
| CN109191403A (en) | Image processing method and device, electronic equipment, computer readable storage medium | |
| Tominaga et al. | Spectral imaging by synchronizing capture and illumination | |
| CN110319790A (en) | A kind of terrible imaging method of full color calculating based on colored speckle field | |
| Koskinen et al. | Reverse imaging pipeline for raw RGB image augmentation | |
| Wenger et al. | Optimizing color matching in a lighting reproduction system for complex subject and illuminant spectra | |
| Catrysse et al. | Comparative analysis of color architectures for image sensors | |
| JP4174707B2 (en) | Spectroscopic measurement system, color reproduction system | |
| CN109923398B (en) | Color imaging with discrete narrowband simultaneous illumination | |
| Rice et al. | Hyperspectral image projector applications | |
| Lam et al. | Spectral Imaging Using Basis Lights. | |
| CN112200825B (en) | Space target luminosity inversion method based on Bell coding image | |
| JP2006090897A (en) | Spectral reflectance estimating system using two kinds of light sources | |
| Tominaga | CIC@ 20: multispectral imaging | |
| CN110332991B (en) | Spectrum reconstruction method and device and electronic equipment | |
| CN106508049B (en) | A kind of process of infrared background view data and conversion method | |
| Berns et al. | Modifications of a sinarback 54 digital camera for spectral and high-accuracy colorimetric imaging: simulations and experiments | |
| CN105744267A (en) | Digital camera parameter varying tristimulus values acquisition method based on homogeneous polynomiald |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |