research article colorimetric analysis using scene ...downloads.hindawi.com › journals › js ›...

Research ArticleColorimetric Analysis Using Scene-AdaptiveColor Conversion Matrix of Calibrated CIS

Sung-Hak Lee

School of Electronics Engineering Kyungpook National University 80 Daehakro Bukgu Daegu 702-701 Republic of Korea

Correspondence should be addressed to Sung-Hak Lee shak2eeknuackr

Received 14 July 2016 Accepted 6 September 2016

Academic Editor Yasuko Y Maruo

Copyright copy 2016 Sung-Hak LeeThis is an open access article distributed under theCreative CommonsAttribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

The RGB signals of different CISs (color image sensors) do not register the same values for the same viewing scene owing to theirdifferent spectral sensitivity and white balance mechanisms Thus CISs must be characterized based on CIE standard observationfor colorimetric purposes One general method for characterizing CISs is least square polynomial modeling to derive thecolorimetric transfer matrix between RGB outputs and CIE 119883119884119885 tristimulus inputs However the transfer matrix that is obtainedunder the standard CIE illumination is unable to estimate various conditions of a CIS that is operated under various illuminationswith varying chromaticity and luminance Therefore repeated experiments are necessary to obtain accurate colorimetric analysisresults This paper presents a scene-adaptive colorimetric analysis method using images captured by a general consumer cameraunder various environments

1 Introduction

Generally the RGB outputs generated by CISs (color imagesensors) are device-dependent This means that the RGBsignals do not correspond to device-independent tristimulusbased on CIE CMFs (color matching functions) [1] This isbecause the spectral sensitivity of CISs is not matched tothe standard CMFs and the characteristics of color sensorsthat are utilized in consumer cameras usually depend on thesurrounding environment when capturing a scene [2]There-fore it is necessary to determine the transfer characteristicsdefining the relationship between RGB signals and standardcolor stimuli for post-color matching and enhancementprocesses such as color reproduction color correction andHDR (high dynamic range) rendering to facilitate accuratescene rendering in general display devices [3ndash8]

The derivation of a transform relationship between RGBsignals and CIE 119883119884119885 stimulus is known as CIS characteri-zation [9] and it can be performed by spectral-sensitivity-based and color-target-based methods The color-target-basedmethod uses reference colors obtained from their119883119884119885values on a color chart and thus is relatively simple and practi-cal when compared to the spectral-sensitivity-based method

that requires spectral analysis of the camera The polynomialregression method based on least squares fitting has beenwidely adopted by many color researchers for calculationof the transfer matrix from the captured RGB values andtheir 119883119884119885 values [9ndash13] However camera characterizationunder the specific standard illuminant does not providegood estimations for the different ambient light conditions inwhich an image is captured Accordingly for more accuratecolor analysis flexible and rigorous experiments in variouswhite balance conditions should be performed

Generally most color images from CISs should be cal-ibrated chromatically to remove specific color cast Thechanged transfer matrix for any other white balance condi-tion can be obtained based on the surrounding illuminationand the phosphor primaries of a camera [12 13] Howevercolor measurement performance can be enhanced using arecalculated transfer matrix based on illuminant estimationtechniques Commonly the illuminant estimation is referredto as white point estimation It is assumed that the CIS is thencalibrated to produce similar RGB values for white under anyilluminant However it is uncertain which illuminant will beused and how similar the RGB values will be to each otherIn theory color correction with unknown white balance is

Hindawi Publishing CorporationJournal of SensorsVolume 2016 Article ID 6731572 7 pageshttpdxdoiorg10115520166731572

2 Journal of Sensors

not problematic because the calibrating coefficients usedto scale the sensor parameters are absorbed into the colortransformation required for the color correction [14]

In this paper a novel colorimetric analysis method is pro-posed using images captured by a consumer digital cameraIn general cases it is hard to estimate the internal calibrationstate of the camera sensor and surrounding illumination Forthis reason without the illumination estimation process thecolor conversion function for various RGB levels is derivedusing only two images captured on the AWB (auto whitebalance) mode and the preset mode The difference of thetwo images is used for the internal calibration process thatcorrects the color balance It is possible to estimate the colorconversion of the image sensor through the color transferfunction In simulation results performance comparisons aremade with the characterized cameras based on the singlematrix and themultimatricesThis demonstrates that the pro-posed camera characterization method performs effectivelyrelative to previous methods

2 Image Calibration

In handling captured images an important issue concerns thecolor balance status of the image sensor In most consumerdigital cameras the sensor color channels are calibratedaccording to sensor response to the intensity and wavelengthof the sceneThe AWB techniques find the optimal color bal-anced condition for varying surroundings by using variousillumination estimation methods such as grey world whitepatch based neural network based and bootstrapping meth-ods [14]Moreover conventional digital cameras adopt a kneecurve for dynamic range compression and the nonsaturatedhighlight range [15]

Figure 1 shows the difference between the sensorresponses of Nikon and Canon cameras Both cameras arebalanced for identical illumination of A and D65 illuminantsfor the same scene This allows a comparison between differ-ent digital cameras as regards color balance transformationSensor responses are plotted in the relative RGB space andshow how the RGB values are calibrated between white bal-anced and unbalanced images For some color patchesthe RGB values are synthesized with the different sensorcharacteristics of two cameras It can be seen that the RGBchannel gains are controlled differently for different signallevels and that the RGB responses vary significantly eventhough light conditions are consistent These are affectedby the camera-specific color constancy algorithms In otherwords the color sensors of two cameras perform differentlyeven for the case of color balanced for the same illuminationThe internal calibrating parameters are absorbed into thecolor transformation and cause nonlinear transformationowing to the gain gamma and knee control properties

3 Colorimetric Analysis

Color imaging systems usingCISs can be characterized by thecolorimetric method The resulting camera characterizationgenerates a transfer matrix for the colorimetric analysis of

captured images The color-target-based method uses refer-ence colors and their corresponding119883119884119885 values to determinean approximated linear transformation between CIE 119883119884119885values and camera RGB signalsThe transfer characterizationmatrix is derived through polynomial regression based onmeasurements of known color samples The color target-based characterization using polynomial regression is gener-ally adopted owing to its simplicity and effectiveness

The transformation processing between119883119884119885 tristimulusfor 119873 color test targets and corresponding camera RGB sig-nals is shown in the following

[[[1198771 1198772 sdot sdot sdot 1198771198731198661 1198662 sdot sdot sdot 1198661198731198611 1198612 sdot sdot sdot 119861119873

]]] = [119872c] [[[1198831 1198832 sdot sdot sdot 1198831198731198841 1198842 sdot sdot sdot 1198841198731198851 1198852 sdot sdot sdot 119885119873

]]]

[119872c] = [[[1198771 1198772 sdot sdot sdot 1198771198731198661 1198662 sdot sdot sdot 1198661198731198611 1198612 sdot sdot sdot 119861119873

]]][[[1198831 1198832 sdot sdot sdot 1198831198731198841 1198842 sdot sdot sdot 1198841198731198851 1198852 sdot sdot sdot 119885119873

]]]119879

sdot ([[[1198831 1198832 sdot sdot sdot 1198831198731198841 1198842 sdot sdot sdot 1198841198731198851 1198852 sdot sdot sdot 119885119873


]]]119879)minus1

(1)

The camera characterization matrix 119872c is determined tominimize the color differences over all test targets Because 3times119873 RGB values should correspond to 3 times 119873 119883119884119885 tristimulusthrough a 3 times 3 matrix the 3 times 3 119872c matrix is derived byusing the least squares method [16]

Nevertheless this method is accurate only under certainstandard illuminants because the characteristics of CISs aredependent on the surrounding illumination In addition theinternal AWB process controls the variations of chromaticityand intensity for correct imaging The cameras have differ-ent colorimetric characteristics according to white balanceconditions Accordingly the transfer matrix should be recal-culated for these inconsistent cases and thus illumination-adaptive measuring methods have been proposed

A recently proposedmethod uses the estimated phosphorprimaries and reference white points to calculate the tristim-ulus constant matrix for certain illumination conditions [12]However this requires phosphor chromaticity informationfor the RGB and white reference to be previously measuredAn alternative method is color measurement based on mul-ticharacterization and illuminant estimation [13] First therepresentative camera transfer matrix is measured for AD50 and D65 illuminants Then the CCT (correlated colortemperature) of an illuminant is estimated to select thecorrect representative transfer matrix for the illuminantFinally the tristimulus weighting factors from ALC (autoluminance control) functions are modified to control signalgains for sustained luminance levels This enables moreprecise estimation results through the selective119872c

Journal of Sensors 3

R

G

B

Approximation

Nikon CIS under A illuminant

Cannon CIS under A illuminant

AWB RG

B ou

tput

(nor

mal

ized

)AW

B RG

B ou

tput

(nor

mal

ized

)

R

G

B

Approximation

00

02

04

06

08

10

02 04 06 08 1000RGB input (normalized)


00

02

04

06

08

10

(a)

AWB RG

B ou

tput

(nor

mal

ized

)

Nikon CIS under D65 illuminant

Cannon CIS under D65 illuminant

AWB RG

B ou

tput

(nor

mal

ized

)

R

G

B

Approximation

R

G

B

Approximation


00

02

04

06

08

10

00

02

04

06

08

10


(b)

Figure 1 Comparisons of relative RGB transform relationships obtained from color balanced images by two different cameras for the sameilluminant (a) under illuminant A (b) under illuminant D65

4 Colorimetric Analysis Based on aCalibrated Color Conversion Matrix

The characteristics and calibration of CISs are dependent onthe surrounding illuminants To obtain a precise camera char-acteristic that changes under different ambient conditions itis necessary to know calibrating parameters such as the statusof the ALC registers and balanced white points However thecharacteristics of balanced images and the sensormechanismare unknown for consumer cameras thus it is very difficultto find out the condition of the calibrated sensors in an arbi-trary ambient situation The previous analytical methods areapplicable only under restricted measuring circumstancesin which it is possible to understand the sensorrsquos signalprocessing by interfacing with the internal registers Thusin this section an ambient-independent method is proposedfor enhanced colorimetric measurements It approaches theresulting color transformation based on the characterizations

for the uncalibrated and calibrated images in each preset andAWB mode

First RGB input signals are converted to normalized val-ues that are compensated linearly with the inverse of the cam-era gamma

119877119873 = ( 1198772119899 minus 1)1120574119877 119866119873 = ( 1198662119899 minus 1)1120574119866 119861119873 = ( 1198612119899 minus 1)1120574119861

(2)

where 119877119873119866119873119861119873 are the gamma compensated and normal-ized outputs from the RGB inputs 119899 is the bit-depth (119899 = 8)and 120574119877120574119866120574119861 are gamma values for each RGB channel


Then the camera characterization for the transfer func-tion should be performed under the reference illuminantThecolor transfer matrix is calculated for a camera set by thepreset mode Herein the 5000K illuminant is adopted forthe reference and RGB channel gains are set internally suchthat RGB outputs are in a ratio of 1 1 1 for white and greysamples under that illuminant Thus the preset mode meansthe reference mode for the D50 illuminant The transfercharacterization of a CIS can be conducted using polynomialregression with the least squares method For the referenceilluminant 119883119884119885 tristimulus are estimated from the derived3 times 3 matrix (119872cD50) and RGB signals in the preset mode asper the following

[[[119883119884119885]]] = [119872cD50]minus1 [[[

119877119873119866119873119861119873]]] (3)

However for the arbitrary illuminant condition the singlecharacterization does not perform an adequate estimationof the reflected color signal The transfer matrix should bemodified to correspond to the changed surroundings

As shown in Figure 1 the calibrated images representthe converted color balances and it is possible to find theweight of primaries and channel gains by using the differ-ence between the uncalibrated and calibrated images Thedifference between the uncalibrated image in the preset (orreference)mode and the calibrated image in the specificAWBcondition can be used to determine a color conversion fordifferent signal levels The transfer gains of each 119877 119866 and 119861for the lower part (119862lg) and upper part (119862ug) are calculatedapproximately as per the following

119862lg [119909] =

low AWBlow reffor 119909 isin 119877119873 | 119909 lt 119909clow AWBlow reffor 119909 isin 119866119873 | 119909 lt 119909clow AWBlow reffor 119909 isin 119861119873 | 119909 lt 119909c

119862ug [119909] =

up AWBup reffor 119909 isin 119877119873 | 119909 ge 119909cup AWBup reffor 119909 isin 119866119873 | 119909 ge 119909cup AWBup reffor 119909 isin 119861119873 | 119909 ge 119909c

(4)

where ldquo rdquo denotes the average value subscript ldquolow AWBrdquoand ldquolow refrdquo refer to the lower part in theAWBmode and thereference mode respectively 119909c is the average value of totalreference mode inputs and ldquoup AWBrdquo and ldquoup refrdquo refer tothe upper part in eachmodeThe calculated data are sampledat a rate of 10 1 to reduce computing burden

According to the separated signals the approxima-tion function for transfer gains consists of two parts The

Lower gain

Upper gain

Fitted WB function

Lower part Upper part

y(A

WB

outp

ut)

Cug

fAWB[x]

ymax

xmaxxlc xc xuc

Clg middot xlc

Cug middot xuc

Clg

000

025

050

075

100

025 050 075 100000x (reference input)

Color samples

Figure 2 Illustration of the estimation of the channel transfer func-tion using lower and upper color gains 119862lg and 119862ug

bi-segmental channel transfer function is derived to reflectthe nonlinear transformation property as per the following

119891119905 (119909) = 119862lg [119909] if 119909 lt 119909lc119891AWB [119909]119909 if 119909 ge 119909lc119891AWB [119909] = 1199100 + 119886119909 + 1198871199092

[[[1199100119886119887 ]]] =

[[[[1 119909max 119909max

21 119909lc 119909lc21 119909uc 119909uc2]]]]minus1 [[[[

119910max119862lg [119909] 119909lc119862ug [119909] 119909uc]]]]

(5)

where 119891AWB is the estimated AWB output 119909lc is the averagevalue of reference values less than 119909c 119909uc is the average valueof reference values more than 119909c and 119909max and 119910max are themaximum values in the reference and AWB modes respec-tively

The value of 119891AWB has been estimated nonlinearly usinga quadratic function for the estimation of calibrated outputscorresponding to the region of reference inputs more than119909lc while a transfer function has been derived linearly forthe relatively narrow region of lower values less than 119909lcThe construction of the channel transfer function using bi-segmental color gains 119862lg and 119862ug is illustrated in Figure 2

Then the new color transfer matrix of the CIS is recon-structed using a color conversion matrix (119872CC) between theuncalibrated and calibrated images as per (6) Finally thetristimulus of color objects in a scene can be estimated usingthe generated transfer matrix

[[[[119883101584011988410158401198851015840]]]] = [119872CC times119872cD50]minus1 [[[

119877119873119860119866119873119860119861119873119860]]]

119872CC = [[[[119891119905 (119877119873119877) 0 00 119891119905 (119866119873119877) 00 0 119891119905 (119861119873119877)

]]]] (6)


Table 1 Comparisons of estimation performance among single-119872c multi-119872c and adaptive-119872c in color differences

Test patchesColor difference Δ1199061015840V1015840

Single-119872c (D50) Multi-119872c Adaptive-119872cD50 D65 A D50 D65 A D50 D65 A

Red 00183 00086 00354 00183 00111 00126 00063 00124 00229Green 00061 00046 00208 00061 00027 00019 00031 00073 00024Blue 00078 00570 00450 00078 00376 00086 00268 00110 00174Cyan 00105 00182 00174 00105 00113 00212 00179 00105 00379Magenta 00092 00226 00456 00092 00151 00321 00142 00044 00277Yellow 00101 00160 00184 00101 00202 00198 00041 00090 00045White 00117 00174 00273 00117 00148 00210 00098 00067 00003Average 00105 00206 00300 00105 00161 00167 00117 00088 00162Model average 00204 00145 00122

Table 2 Comparisons of illuminant estimation among single-119872cmulti-119872c and adaptive-119872c in color differences

CCT of illuminant Color difference Δ1199061015840V1015840Single-119872c (D50) Multi-119872c Adaptive-119872c

D50 00117 00117 00098D65 00174 00148 00067A (2800K) 00273 00210 00003Average 00188 00158 00056

where 119877119873119877119866119873119860119861119873119860 are the calibrated inputs in the AWBmode and 119877119873119877119866119873119877119861119873119877 are the uncalibrated inputs in thereference mode 119883101584011988410158401198851015840 are values for general nonstandardilluminant environments

When the illuminant changes the color conversionmatrix controls 3 channel gains according to the image cal-ibration by AWB processing The proposed colorimetricanalysis flow is shown in Figure 3(b) in comparison withthe multi-119872c method of Figure 3(a) The multi-119872c methodrequires finding characteristic transfer matrices under var-ious standard illuminations and estimating the illuminantOn the other hand the adaptive-119872c gives estimated resultsfor illuminants and color information by obtaining the colorconversion matrix using images captured in two differentmodes without the illuminant estimation and the AWBinformation of the image sensor

5 Simulation Results

Theproposed algorithmhas been tested on images containingMacbeth 24 color samples and various objects under therepresentative D50 D65 and A illuminants using a Macbethlighting booth A schematic diagram of the camera character-ization and color estimation is shown in Figure 4The viewingconditions considered for the verification of the methodare restricted within the range of indoor environments andthree illuminants A CMOS CIS (TOSHIBA TCM8230MD)module including a control interface was used for the exper-iments To evaluate the multi-119872c method compared with theproposed method first the illuminant was estimated usinga single characterization matrix (D50) for higher luminancepixels corresponding to the upper 10 of 119884 signalsThen the

most closematching of the three illuminants was selected andthe corresponding characterizationmatrix was used for colorestimations For the consideration of the preset mode theinternal parameters of the camera were fixed so that the RGBoutputs maintained the white balance state under the 5000Killuminant

Tables 1 2 and 3 show the relative performance of thecolor and illuminant estimations in color differences betweenthe measured and predicted values and a comparison of theprocessing steps between multi-119872c and adaptive-119872c Theaverage errors were computed in CIE 1198711199061015840V1015840 color spaceComparisons of average color differences for color samplesand illuminants for the three methods are shown in Tables 1and 2 respectively The results show that the adaptive-119872cmethod performs better than the multi-119872c method by 16and 65 for the estimation of color samples and illuminantsrespectively The meaningful results are shown in color esti-mations under illuminantD65 andwhite estimations forD50D65 andA illuminants In comparisonwith the conventionalmulticharacterization for typical illuminants through theilluminant estimation process the proposedmethod requiresonly a single characterization no ALC information and no119872c selection based on illuminant estimation however it doesrequire acquiring two-mode images

In Table 3 the multi-119872c method requires finding thecharacteristic transfer matrices for various standard illu-minations Additionally the illuminant estimation for 119872cselection using pixel samples of higher levels is inaccurateHowever the adaptive-119872c demonstrates satisfactory resultsfor illuminants and color information by obtaining the colorconversion matrix using images captured in two differentmodes without any other interfacing with the CIS

6 Conclusions

In color imaging systems colorimetric measuring is neces-sary to correctly reproduce color images This is generallyperformed by techniques such as chromatic adaptation trans-formation that requires absolute luminance and chrominanceinformation

It is also indispensable to estimate illuminants and objectcolors for the correction of degraded effects by tonal map-ping and color distortion in rendering high dynamic range


Estimated XYZ

RGB

Manual CISconditions

Illuminationestimation

CCT

Viewingscene

AWB off

gamma off

ALC status

RGB gain times1 RGBrarr XYZ

RGBrarr XYZ

from selected McMc selector

from Mc(D50)

Mc(A)

Mc(D50)

Mc(D65)

(a)

Channel gain estimation

Estimated XYZ

AWB RGB

Manual CISconditionsViewing

sceneAWB on

amppreset

mode (D50)

from color conversion matrix

Preset RGB Cug

Clg

RGBrarr XYZ

Lower gain (Clg [x])

Upper gain (Cug [x])amp preset Mc

[MCC times McD50 ]minus1

(b)

Figure 3 Block diagrams of color analysis processes (a) the multi-119872c method and (b) the scene-adaptive-119872c method

Output value RGB Tristimulus value XYZ

TOSHIBA TCM8230MD

Capture amp camera controller

Target colors amp illuminants (D50 D65 A)

Camera control SW

GretagMacbethcolor checker chart

Measurement

Chroma meterCS-100A

Camera characterization amp color estimation

⟨D50⟩

⟨D65⟩ ⟨A⟩

Figure 4 Schematic of the camera characterization and color estimation for the evaluation


Table 3 Comparisons between multi-119872c and adaptive-119872c in terms of processing steps and results

Multi-119872c based on illuminant estimation Adaptive-119872c based on color conversion estimation

Processing steps

(1) Set the CIS AWB off manual RGB gain (1 1 1)gamma correction off(2) Characterizing a CIS for 3 standard illuminants (AD50 D65)(3) Illuminant estimation (white point)(4) Selecting the transfer matrix(5) Reading ALC registers(6) Color estimation

(1) Characterizing CIS for only a standard illuminant(D50)(2) Capturing with AWB on(3) Capturing in preset mode (D50)(4) Calculation of color conversion matrix(5) Color estimation

Δ1199061015840V1015840Color samples 00142 00133Illuminants (white

samples) 00158 00056

scenes on low dynamic range displays However it is dif-ficult to obtain accurate colorimetric measurement becausethe AE and AWB of cameras operate according to theexposed environment The status of sensor calibration isunknown because it differs in many imaging systems Ingeneral color conversion techniques have been developed toconvert between different illuminants based strictly on thephysical quantities and thus the measured results depend onthe chosen illuminant

In this paper to solve this problem the transfer matrixaccording to white-balance conditions is estimated but doesnot use the internal CIS register values The differencebetween the uncalibrated and calibrated images is solely usedto derive the color conversion matrix This method of coloranalysis can be consistent with the assumption of knowncolor-processing and illuminants Experimental results showthe proposed method is valid in terms of measuring per-formance The prediction of luminance and chromaticity ofscenes is applicable to CISs in automatic systems respondingto various environments

Competing Interests

The author declares that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea (NRF) funded by the Ministry of Education (NRF-2015R1D1A1A01059929)

References

[1] CIE Publication No 152 Colorimetry Central Bureau of theCIE Vienna Austria 2nd edition 1986

[2] S Hullfish and J FowlerColor Correction for Video Focal Press2nd edition 2008

[3] P-C Hung ldquoColorimetric calibration in electronic imagingdevices using a look-up-table model and interpolationsrdquo Jour-nal of Electronic Imaging vol 2 no 1 pp 53ndash61 1993

[4] M D Fairchild Color Appearance Models John Wiley amp SonsNew York NY USA 2nd edition 2005

[5] G W Larson H Rushmeier and C Piatko ldquoA visibility match-ing tone reproduction operator for high dynamic range scenesrdquoIEEE Transactions on Visualization and Computer Graphics vol3 no 4 pp 291ndash306 1997

[6] E ReinhardM Stark P Shirley and J Ferwerda ldquoPhotographictone reproduction for digital imagesrdquo ACM transactions onGraphics vol 21 no 3 pp 267ndash276 2002

[7] N Moroney M D Fairchild R W G Hunt C Li M R Luoand T Newman ldquoThe CIECAM02 color appearance modelrdquo inProceedings of the ISampTrsquos Color and Imaging Conference (CICrsquo02) pp 23ndash27 Scottsdale Ariz USA 2002

[8] R Mantiuk R Mantiuk A Tomaszewska and W HeidrichldquoColor correction for tone mappingrdquo Computer GraphicsForum vol 28 no 2 pp 193ndash202 2009

[9] GHongM R Luo and P A Rhodes ldquoA study of digital cameracolorimetric characterization based on polynomial modelingrdquoColor Research and Application vol 26 no 1 pp 76ndash84 2001

[10] T Johnson ldquoMethods for characterizing colour scanners anddigital camerasrdquo Displays vol 16 no 4 pp 183ndash191 1996

[11] H R KangTheColor Technology for Electronic Imaging DevicesSPIE Optical Engineering Press Bellingham Wash USA 1997

[12] E-S Kim S-H Lee S-W Jang and K-I Sohng ldquoAdaptivecolorimetric characterization of camera for the variation ofwhite balancerdquo IEICE Transactions on Electronics vol E88-Cno 11 pp 2086ndash2089 2005

[13] S-H Lee J-H Lee and K-I Sohng ldquoAn illumination-adaptivecolorimetric measurement using color image sensorrdquo IEICETransactions on Electronics vol 91 no 10 pp 1608ndash1610 2008

[14] V C Cardei B Funt and K Barndard ldquoWhite point estimationfor uncalibrated imagesrdquo in Proceedings of the ISampT 7th ColorImaging Conference pp 97ndash100 1999

[15] Y Monobe H Yamashita T Kurosawa and H KoteraldquoDynamic range compression preserving local image contrastfor digital video camerardquo IEEE Transactions on ConsumerElectronics vol 51 no 1 pp 1ndash10 2005

[16] D G Zill andM R CullenAdvanced EngineeringMathematicsvol 2nd Jones and Bartlett Publishers International 1999

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



not problematic because the calibrating coefficients usedto scale the sensor parameters are absorbed into the colortransformation required for the color correction [14]

In this paper a novel colorimetric analysis method is pro-posed using images captured by a consumer digital cameraIn general cases it is hard to estimate the internal calibrationstate of the camera sensor and surrounding illumination Forthis reason without the illumination estimation process thecolor conversion function for various RGB levels is derivedusing only two images captured on the AWB (auto whitebalance) mode and the preset mode The difference of thetwo images is used for the internal calibration process thatcorrects the color balance It is possible to estimate the colorconversion of the image sensor through the color transferfunction In simulation results performance comparisons aremade with the characterized cameras based on the singlematrix and themultimatricesThis demonstrates that the pro-posed camera characterization method performs effectivelyrelative to previous methods

2 Image Calibration

In handling captured images an important issue concerns thecolor balance status of the image sensor In most consumerdigital cameras the sensor color channels are calibratedaccording to sensor response to the intensity and wavelengthof the sceneThe AWB techniques find the optimal color bal-anced condition for varying surroundings by using variousillumination estimation methods such as grey world whitepatch based neural network based and bootstrapping meth-ods [14]Moreover conventional digital cameras adopt a kneecurve for dynamic range compression and the nonsaturatedhighlight range [15]

Figure 1 shows the difference between the sensorresponses of Nikon and Canon cameras Both cameras arebalanced for identical illumination of A and D65 illuminantsfor the same scene This allows a comparison between differ-ent digital cameras as regards color balance transformationSensor responses are plotted in the relative RGB space andshow how the RGB values are calibrated between white bal-anced and unbalanced images For some color patchesthe RGB values are synthesized with the different sensorcharacteristics of two cameras It can be seen that the RGBchannel gains are controlled differently for different signallevels and that the RGB responses vary significantly eventhough light conditions are consistent These are affectedby the camera-specific color constancy algorithms In otherwords the color sensors of two cameras perform differentlyeven for the case of color balanced for the same illuminationThe internal calibrating parameters are absorbed into thecolor transformation and cause nonlinear transformationowing to the gain gamma and knee control properties

3 Colorimetric Analysis

Color imaging systems usingCISs can be characterized by thecolorimetric method The resulting camera characterizationgenerates a transfer matrix for the colorimetric analysis of

captured images The color-target-based method uses refer-ence colors and their corresponding119883119884119885 values to determinean approximated linear transformation between CIE 119883119884119885values and camera RGB signalsThe transfer characterizationmatrix is derived through polynomial regression based onmeasurements of known color samples The color target-based characterization using polynomial regression is gener-ally adopted owing to its simplicity and effectiveness

The transformation processing between119883119884119885 tristimulusfor 119873 color test targets and corresponding camera RGB sig-nals is shown in the following

[[[1198771 1198772 sdot sdot sdot 1198771198731198661 1198662 sdot sdot sdot 1198661198731198611 1198612 sdot sdot sdot 119861119873

]]] = [119872c] [[[1198831 1198832 sdot sdot sdot 1198831198731198841 1198842 sdot sdot sdot 1198841198731198851 1198852 sdot sdot sdot 119885119873

]]]

[119872c] = [[[1198771 1198772 sdot sdot sdot 1198771198731198661 1198662 sdot sdot sdot 1198661198731198611 1198612 sdot sdot sdot 119861119873


]]]119879

sdot ([[[1198831 1198832 sdot sdot sdot 1198831198731198841 1198842 sdot sdot sdot 1198841198731198851 1198852 sdot sdot sdot 119885119873


]]]119879)minus1

(1)

The camera characterization matrix 119872c is determined tominimize the color differences over all test targets Because 3times119873 RGB values should correspond to 3 times 119873 119883119884119885 tristimulusthrough a 3 times 3 matrix the 3 times 3 119872c matrix is derived byusing the least squares method [16]

Nevertheless this method is accurate only under certainstandard illuminants because the characteristics of CISs aredependent on the surrounding illumination In addition theinternal AWB process controls the variations of chromaticityand intensity for correct imaging The cameras have differ-ent colorimetric characteristics according to white balanceconditions Accordingly the transfer matrix should be recal-culated for these inconsistent cases and thus illumination-adaptive measuring methods have been proposed

A recently proposedmethod uses the estimated phosphorprimaries and reference white points to calculate the tristim-ulus constant matrix for certain illumination conditions [12]However this requires phosphor chromaticity informationfor the RGB and white reference to be previously measuredAn alternative method is color measurement based on mul-ticharacterization and illuminant estimation [13] First therepresentative camera transfer matrix is measured for AD50 and D65 illuminants Then the CCT (correlated colortemperature) of an illuminant is estimated to select thecorrect representative transfer matrix for the illuminantFinally the tristimulus weighting factors from ALC (autoluminance control) functions are modified to control signalgains for sustained luminance levels This enables moreprecise estimation results through the selective119872c


R

G

B

Approximation



AWB RG

B ou

tput

(nor

mal

ized

)AW

B RG

B ou

tput

(nor

mal

ized

)

R

G

B

Approximation

00

02

04

06

08

10



00

02

04

06

08

10

(a)

AWB RG

B ou

tput

(nor

mal

ized

)



AWB RG

B ou

tput

(nor

mal

ized

)

R

G

B

Approximation

R

G

B

Approximation


00

02

04

06

08

10

00

02

04

06

08

10


(b)







(2)




[[[119883119884119885]]] = [119872cD50]minus1 [[[

119877119873119866119873119861119873]]] (3)



119862lg [119909] =


119862ug [119909] =


(4)



Lower gain

Upper gain

Fitted WB function


y(A

WB

outp

ut)

Cug

fAWB[x]

ymax

xmaxxlc xc xuc

Clg middot xlc

Cug middot xuc

Clg

000

025

050

075

100


Color samples




[[[1199100119886119887 ]]] =

[[[[1 119909max 119909max

21 119909lc 119909lc21 119909uc 119909uc2]]]]minus1 [[[[

119910max119862lg [119909] 119909lc119862ug [119909] 119909uc]]]]

(5)




[[[[119883101584011988410158401198851015840]]]] = [119872CC times119872cD50]minus1 [[[

119877119873119860119866119873119860119861119873119860]]]

119872CC = [[[[119891119905 (119877119873119877) 0 00 119891119905 (119866119873119877) 00 0 119891119905 (119861119873119877)

]]]] (6)








D50 00117 00117 00098D65 00174 00148 00067A (2800K) 00273 00210 00003Average 00188 00158 00056








6 Conclusions




Estimated XYZ

RGB



CCT

Viewingscene

AWB off

gamma off

ALC status


RGBrarr XYZ


from Mc(D50)

Mc(A)

Mc(D50)

Mc(D65)

(a)


Estimated XYZ

AWB RGB


sceneAWB on

amppreset

mode (D50)


Preset RGB Cug

Clg

RGBrarr XYZ




(b)



TOSHIBA TCM8230MD



Camera control SW


Measurement

Chroma meterCS-100A


⟨D50⟩

⟨D65⟩ ⟨A⟩





Processing steps




samples) 00158 00056



Competing Interests


Acknowledgments


References



















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










R

G

B

Approximation



AWB RG

B ou

tput

(nor

mal

ized

)AW

B RG

B ou

tput

(nor

mal

ized

)

R

G

B

Approximation

00

02

04

06

08

10



00

02

04

06

08

10

(a)

AWB RG

B ou

tput

(nor

mal

ized

)



AWB RG

B ou

tput

(nor

mal

ized

)

R

G

B

Approximation

R

G

B

Approximation


00

02

04

06

08

10

00

02

04

06

08

10


(b)







(2)




[[[119883119884119885]]] = [119872cD50]minus1 [[[

119877119873119866119873119861119873]]] (3)



119862lg [119909] =


119862ug [119909] =


(4)



Lower gain

Upper gain

Fitted WB function


y(A

WB

outp

ut)

Cug

fAWB[x]

ymax

xmaxxlc xc xuc

Clg middot xlc

Cug middot xuc

Clg

000

025

050

075

100


Color samples




[[[1199100119886119887 ]]] =

[[[[1 119909max 119909max

21 119909lc 119909lc21 119909uc 119909uc2]]]]minus1 [[[[

119910max119862lg [119909] 119909lc119862ug [119909] 119909uc]]]]

(5)




[[[[119883101584011988410158401198851015840]]]] = [119872CC times119872cD50]minus1 [[[

119877119873119860119866119873119860119861119873119860]]]

119872CC = [[[[119891119905 (119877119873119877) 0 00 119891119905 (119866119873119877) 00 0 119891119905 (119861119873119877)

]]]] (6)








D50 00117 00117 00098D65 00174 00148 00067A (2800K) 00273 00210 00003Average 00188 00158 00056








6 Conclusions




Estimated XYZ

RGB



CCT

Viewingscene

AWB off

gamma off

ALC status


RGBrarr XYZ


from Mc(D50)

Mc(A)

Mc(D50)

Mc(D65)

(a)


Estimated XYZ

AWB RGB


sceneAWB on

amppreset

mode (D50)


Preset RGB Cug

Clg

RGBrarr XYZ




(b)



TOSHIBA TCM8230MD



Camera control SW


Measurement

Chroma meterCS-100A


⟨D50⟩

⟨D65⟩ ⟨A⟩





Processing steps




samples) 00158 00056



Competing Interests


Acknowledgments


References



















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











[[[119883119884119885]]] = [119872cD50]minus1 [[[

119877119873119866119873119861119873]]] (3)



119862lg [119909] =


119862ug [119909] =


(4)



Lower gain

Upper gain

Fitted WB function


y(A

WB

outp

ut)

Cug

fAWB[x]

ymax

xmaxxlc xc xuc

Clg middot xlc

Cug middot xuc

Clg

000

025

050

075

100


Color samples




[[[1199100119886119887 ]]] =

[[[[1 119909max 119909max

21 119909lc 119909lc21 119909uc 119909uc2]]]]minus1 [[[[

119910max119862lg [119909] 119909lc119862ug [119909] 119909uc]]]]

(5)




[[[[119883101584011988410158401198851015840]]]] = [119872CC times119872cD50]minus1 [[[

119877119873119860119866119873119860119861119873119860]]]

119872CC = [[[[119891119905 (119877119873119877) 0 00 119891119905 (119866119873119877) 00 0 119891119905 (119861119873119877)

]]]] (6)








D50 00117 00117 00098D65 00174 00148 00067A (2800K) 00273 00210 00003Average 00188 00158 00056








6 Conclusions




Estimated XYZ

RGB



CCT

Viewingscene

AWB off

gamma off

ALC status


RGBrarr XYZ


from Mc(D50)

Mc(A)

Mc(D50)

Mc(D65)

(a)


Estimated XYZ

AWB RGB


sceneAWB on

amppreset

mode (D50)


Preset RGB Cug

Clg

RGBrarr XYZ




(b)



TOSHIBA TCM8230MD



Camera control SW


Measurement

Chroma meterCS-100A


⟨D50⟩

⟨D65⟩ ⟨A⟩





Processing steps




samples) 00158 00056



Competing Interests


Acknowledgments


References



















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation
















D50 00117 00117 00098D65 00174 00148 00067A (2800K) 00273 00210 00003Average 00188 00158 00056








6 Conclusions




Estimated XYZ

RGB



CCT

Viewingscene

AWB off

gamma off

ALC status


RGBrarr XYZ


from Mc(D50)

Mc(A)

Mc(D50)

Mc(D65)

(a)


Estimated XYZ

AWB RGB


sceneAWB on

amppreset

mode (D50)


Preset RGB Cug

Clg

RGBrarr XYZ




(b)



TOSHIBA TCM8230MD



Camera control SW


Measurement

Chroma meterCS-100A


⟨D50⟩

⟨D65⟩ ⟨A⟩





Processing steps




samples) 00158 00056



Competing Interests


Acknowledgments


References



















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Estimated XYZ

RGB



CCT

Viewingscene

AWB off

gamma off

ALC status


RGBrarr XYZ


from Mc(D50)

Mc(A)

Mc(D50)

Mc(D65)

(a)


Estimated XYZ

AWB RGB


sceneAWB on

amppreset

mode (D50)


Preset RGB Cug

Clg

RGBrarr XYZ




(b)



TOSHIBA TCM8230MD



Camera control SW


Measurement

Chroma meterCS-100A


⟨D50⟩

⟨D65⟩ ⟨A⟩





Processing steps




samples) 00158 00056



Competing Interests


Acknowledgments


References



















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












Processing steps




samples) 00158 00056



Competing Interests


Acknowledgments


References



















RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









research article colorimetric analysis using scene ...downloads.hindawi.com › journals › js ›...

Documents