Studying for the Enhancement Captured Images at Different Lightness Directions and Levels Distribution Based on Using Histogram Equalization Method

Transcription

1 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 Studyig for the Ehacemet Captured Images at Differet Lightess Directios ad Levels Distributio Based o Usig Histogram Equalizatio ethod Awar H.. Al-Saleh *, Salema S.S. Zuheri, Ali A.D. Al-Zuky 3 Departmet of Computer Sciece, College of Sciece, Uiversity of Al- ustasariya, Baghdad, Iraq Departmet of Cliical Laboratory Sciece, College of Pharmacy, Uiversity of Baghdad, Baghdad, Iraq 3 Departmet of Physics, College of Scieces, Uiversity of Al- ustasariya, Baghdad, Iraq Abstract I this research we study the effect of the light distributio o the quality of captured images uder differet amout of lightess coditios the ehacig the captured images by the imagig system usig a adaptive histogram equalizatio method ad Lee method depedig o the space YIQ depedig o lightess compoet where use the reverse of this space to the basic RGB color space, the corrected usig the mathematical model based o huma visio system. The results were aalyzed ad compute the quality of the ehacemet images by usig various statistical criteria based o the mea ad stadard deviatio. Adaptive histogram equalizatio techique gives a high quality for the ehaced images for differet lightess coditios. 3 * 3 YIQ RGB... Itroductio Image cotrast ehacemet plays a essetial role i digital image processig []. The objective of image ehacemet techique is to improve the image quality so that the processed image is better tha the origial image for a specific applicatio or set of objectives []. Image quality plays a importat role i various image processig applicatios [3]. The quality of the images could be affected by several factors. Oe of the most importat factors is the illumiatio, which could be atural or artificial. Ay weak i this factor will lead to a weak images quality, which will eed

2 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 to be ehaced through oe or several image ehacemet correctio [4]. To improve the image cotrast, umerous ehacemet techiques have bee proposed. Oe of the covetioal methods adopted is the Adaptive Histogram Equalizatio ( AHE ) techique[]. This method is simple ad efficiet whe comparig with other method. Adaptive Histogram equalizatio tries to distribute more cetral gray level, extesively[5]. There are may studies that focused o improvig digital images from where to improve the cotrast ad image aalysis to determied its quality, ad the most importat of these studies are:- Eli PeLi (99): proposed method for cotrast ehacemet depedet o the physical cotrast of simple images such as siusoidal gratigs or a sigle patch of light o a uiform backgroud which is well defied ad agree with the perceived cotrast [6]. D. J. Jobso et.al. (996): itroduced ew algorithm to improve the brightess, cotrast ad sharpess of a image. It performs a oliear spatial/spectral trasform that provides simultaeous dyamic rage compressio, ad i 997 they compared this method with aother ehacemet techique such as histogram equalizatio homomorphic filterig [7]. B.V.Fut et.al. (997): itroduced ivestigatios ito ulti-scale Retiex algorithm approach to image ehacemet to explai the effect of the processig from a theoretical stadpoit. I the same year they modified the multi-scale retiex approach to image ehacemet so that the processig is more justified from a theoretical stadpoit by suggested a ew algorithm with fewer arbitrary parameters that is more flexible [8]. Ali J. Al Dalawy (8): studied the TV- Satellite images of Al-Hurra chael broadcasted o Arabsat, Hotbird ad Nilesat. These image were the same with respect to the type o the three satellites. Aalyzig these images doe statistically by fidig the statistics distributio ad studyig the relatios betwee the mea ad the stadard deviatio of the color compoud (RGB) ad light compoet for the image as whole ad for the extracted homogeeous regios. Also he studied the cotrast of image edges depedig o sobel operator i limitig the edges ad studied the cotrast as fuctio for edge fidig threshold [9]. Salema S. Salma(9) studied the effect of differet lightig operatios i type ad itesity o test images usig differet light sources (tugste ad fluorescet lamp) ad studied the distributio homogeeity of light itesity for the lie partitioed from the middle of white test image width ad height ad focused o the cotrast ratio as a fuctio for light itesity usig a test image with oe half white ad the other half black []. Diaa C. Gil et.al. (): They itroduced cotrast ehacemet methods ad Image quality is evaluated by meas of objective metrics such as itesity cotrast ad brightess error, ad by subjective assessmet. Executio time is also measured. They foud the techique based o histogram modificatio presets a better trade-off cosiderig both aspects []. Statistical Digital Image Properties There are several image properties ca be calculated from image data, the most imported properties (mea µ, stadard deviatio σ of the image or image regios) []. - The ea (µ) Image mea brightess is kow as the mea brightess for the image elemets ad it determied from the followig relatioship [3]:- N µ = I( () N x= x= Or by the followig equitatio:- L = g= µ gp( g) () Where I( is the itesity for the elemet image N is the image size. P(g) is the probability distributio desity i image. L is the umbers of desity levels i image. - Stadard Deviatio (STD) The stadard deviatio represet the mea of variatios of the elemet values with respect to its mea ad it determied from the followig relatioship [4]:- σ = N N x= y= ( I( µ ) (3) Or by the followig equitatio:- N F = I ( (4) N x= y= L = I g p( g)..(5) g= = I..(6) σ µ

3 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 This stadard is a importat criterio to determie the amout of detail i the image. Lee Ehacemet Algorithm Is a algorithm to improve the visual quality of digital images captured uder extremely low or uiform lighteig coditios. That is composed of three mai parts: adaptive lumiace ehacemet, cotrast ehacemet ad color restoratio, this algorithm ca be discus as flowig [5]: a. Adaptive lumiace ehacemet: that is dividig i two steps; first step is lumiace estimatio to obtai by coversio of the lumiace iformatio by usig NTSC (Natioal Televisio Stadards Committee) color space. Itesity values of RGB image ca be obtaied as [6]: I ( =.98r( +.587g( +.4b(...(7) ad the Normalizatio Itesity is: I ( = I( / 55...(8) The image iformatio accordig to huma visio behavior ca be simplified ad formulated as [9]: I ( = L( R(...(9) Where R( is the reflectace ad L( is the illumiatio at each positio (, the lumiace L is assumed to be cotaied i the low frequecy compoet of the image while the reflectace R maily represets the high frequecy compoets of the image. The estimatio of the illumiatio, the Gaussia lowpass filtered result of the itesity image is used. I spatial domai, this process is a D discrete covolutio with a Gaussia kerel which ca be expressed as [5]: I ( = L( c = N m= = I( m, ) F( m+ + y, c)...() I c is image covolutio, The secod step is adaptive dyamic rage compressio of lumiace by usig the trasfer fuctio I' by [5]: (.75z+.5) ( z) I.4( )( ) ' + I z + I I =..() Where z is the image depedet parameter calculated by [9]: for L 5 L 5 z = for 5< L...() for L> 5 This ca be achieved by the followig steps:. Iput color image C(,m,i), i=,,3 (red, gree, blue) compoets.. Trasform color image C(,m) from RGB space to YIQ space ad estimated lightess compoet Y(,m). 3. Normalized lightess compoet I(m,)= Y(,m)/ Calculated Gaussia surrouds fuctio ( x y ) F( y, c) = ke c where k is ormalizatio costat, c, =3, {c=5, c=, c3=4}. 5. Computed covolutio image from I ( = N I( m, ) F( m+ + y, c c m= = 6. Calculated.4 I + ( I ).5 + I T = 7. computed reflectace image from ' E( R ( = 55I (, Ic( p E ( = ( ) I (,P= 8. Output image result from compoet I' I' I' r' = r, g' = g, b' = b I I I. Adaptive Histogram Equalizatio (AHE) Algorithm To ehace captured images uder differet lightess coditios from lightig system usig Adaptive Histogram Equalizatio, this performed by usig YIQ color space ad improve oly the lumiace compoet. The techical steps as follows:- The first step i this algorithm is trasformig color image from basic RGB color space to YIQ color space, the forward trasform is give by the followig equatios: Y =.99r +.587g +.4b I =.596r.7g.3b (3) Q =.r.53g +.3b Where Y represet lightess compoet, ad I,Q color compoet. I the secod step is trasformig ormalized lightess value usig sigmoid fuctio that is give by: Y = + I I..(4) ) 3

4 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 Where I, Y are represeted the iteral ad exteral lightig fuctios respectively, usig Histogram Equalizatio techique. The third step is applyig HE o modify lightess compoet, the processig lightess compoet Y has bee gotte form this step. Fially iverse trasformatio from YIQ to RGB color space calculated i YpIQ that is give by [5]: r = y +.956i +.6q g=y.7i.647q.. (5) b=y.6i +.73q This ca be achieved by the followig steps:. Iput color image C(,m,i), i=,,3 (red, gree, blue) compoets..trasform color image C(,m) from RGB space to YIQ space ad estimated lightess compoet Y. 3. Normalized Y compoet by: I = Y /5. 4. Trasformed lightess compoet by usig: Y = I + 5. Applied HE o trasformed lightess compoet I gettig processed lightess compoet y. 6. Trasform color image from Y IQ to get out put image. Results ad Discussio I a homogeeous lightig with differet cases lightig, figures () shows the resultig images for differet cases of lightig before applyig ehacemet methods, ad figures(3,4) show the result images after usig the Lee ehacemet method, ad AHE method respectively. V = 6 V = V = 8 V = V = 4 V = 6 V = 8 V = Light Source Camera Captured Image Figure - The Lightig System Figure () shows lightig system, which cosists of a dark box cotaiig i oe of its sides the light source (lamp from the right aother from the left) to get a regular ad irregular lightig. At the same side i the bottom of the lightig source there is a hole to put the camera. O the other side is placed the image to be captured uder differet lightig coditios. All images i this study are take by Soy Digital Camera, these captured images are of the (JPEG) format ad with differet itesities lightig depedig o the voltage hagig o the light source. We capture images, usig V = V = 4 V = 5 Figure - The Origi Images Before Usig Ehacemet Techiques. 4

5 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 V = 6 V = 8 V = 6 V = 8 V = V = V = V = V = 4 V = 6 V = 4 V = 6 V = 8 V = V = 8 V = V = V = 4 V = V = 4 V = 5 V = 5 Figure 3- The Ehacemet Images Usig Lee Techique. Figure 4- The Ehacemet Images Usig AHE Techique. 5

6 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 To assess the efficiecy of techiques that is used i image ehacemet i this study, by usig differet lightig, we calculate the statistical properties of images the mea (µ) ad the stadard deviatio (STD), where the statistical properties were draw as a fuctio of voltage lightig system before ad after the use of two ehacemet methods, Lee method ad AHE method. Ad the results show i figurers (5,6,7) (a) Lee B L R G 6 Orgi Lee ST L ST R ST G ST B 8 6 S T L R G 9 8 B (a) Orgi (b) Figure 6- The Relatioship Betwee (µ) Ad (STD) Of The Color Compoud (RGB) Ad Light Compoet L For The Images As A Fuctio Of age After Usig Lee Ehacemet ethod. 7 6 S T ST L Adaptive ST R ST G ST B (b) Figure 5- The Relatioship Betwee (µ) Ad (STD) Of The Color Compoud (RGB) Ad Light Compoet L For The Images As A Fuctio Of age Before Usig The Ehacemet ethods. As clear from the figures the best image improve before ehacemet methods i v = 4 volt, where the value of the overall cotrast are higher i this case, ad whe icreasig the voltage more, we ote that the value of the cotrast was decreased although icreasig the light itesity. But after the ehacemet process, we ote that the cotrast value almost equal for all ehaced images. As show i figures (6,7). ST B (a) ST L ST B (b) Adaptive ST G ST R L R G Figure 7- The Relatioship Betwee (µ) Ad (STD) Of The Color Compoud (RGB) Ad Light Compoet L For The Images As A Fuctio Of age After Usig AHE Ehacemet ethod. 6

7 Al-Saleh et.al. Iraqi Joural of Sciece, December, Vol. 53, No. 4, Pp. -7 Coclusio I this study, we improve the Lee ad AHE ehacemet methods based o the YIQ space by processig the lightig compoet oly where the use of reverse coversio of this space to the basic space RGB colors ad the corrected usig the theory based o huma visio system. The result show that the algorithm deal with images color with better results, ad improves the image statistical properties (µ,σ) for images ehacemet are to maitai the geeral attributes of statistical properties of image with differet lightigs accordig the lightig itesity of imagig system. Refereces [] Sharmila R. ad Uma R., "A New Approach To Image Cotrast Ehacemet usig Weighted Threshold Histogram Equalizatio with Improved Switchig edia Filter", Iteratioal Joural of Advaced Egieerig Scieces ad Techologies Vol. 7, Issue o.,pp. 6. [] Jisha T., Eli P., ad Scott A., 3,"Image Ehacemet Usig a Cotrast easure i the Compressed Domai", IEE Sigal Processig Letters, Vol., o.. [3] Brigier B., Richard N., Larabi.-C. ad Feradez-aloige C., 6 "No- Referece Perceptual quality Assessmet of Colour Image", 4th Europea Sigal Processig Coferece (EUSIPCO 6), Florece, Italy, pp. 4-8,. [4] Ebrahim B. 5"Studyig The Effect of Some Image Ehacemet Features o The Accuracy of Close Rage Photogrammetric easuremets Usig CAD Eviromet", Iteratioal Symposium, 6 September October, Torio, Italy. [5] Phathua N., Cheevasuvit F. ad Chitwog S. "Cotrast Ehacemet For iimum ea Brightees Error From Histogram Partitioig". [6] Eli Peli, 99, Cotrast i complex images, J. Opt. Am. A/Vol. 7, No.,. [7] Jobso D. J., Rahma Z., ad Woodell G. A., 997 A Compariso of the ultiscale Retiex With Other Image Ehacemet Techiques, Proceedigs of the IS&T s 5 th Aual Coferece, Cambridge, A,. [8] Barard K. ad Fud B., 997 Aalysis ad Improvemet of ulti-scale Retiex I Proceedigs of the IS&T/SID Fifth Color Imagig Coferece: Color Sciece, Systems ad Applicatios, Scottsdale, pp. -6,. [9] 9- Ali J. Al-Dalawy, 8 A Study of TV Images Quality for Chaels Broadcast Televisio Satellite, aster of Sciece i physics Departmet, Al-ustasiriya Uiversity. [] Salema S. Salma, 9 A Study Lightig Effect i Determiig of Test Image Resolutio, s.c thesis College of Sciece, Al-ustasiriya Uiversity. [] Diaa Gil, Raa Farah, J.. Pierre Laglois, Guillaume-Alexadre Bi, Yvo Savaria, Comparative aalysis of cotrast ehacemet algorithms i surveillace imagig, IEEE, ISCAS, pp ,. [] Ali A. al-zuky, Examiig Cotrast by Chage of Lightess ad Spatial Frequecy, Al ustasiryha J. Sci, V,6,pp.6-6. [3] Ali A. al-zuky, Salma S.S. Zuheri, ad Kathem J. AL-Shujari, Assessmet Techiques for the cotrast of Images have Differet Illumiatio, Al ustasiryha J. Sci. (6): -9. [4] Hala N. Kadom, Ali A. al-zuky, ad Kathem J. AL-Shujari, A Study o Nokia 76 Cell-Phoe Camera Statistics Properties i Differet Lightess, Al ustasiryha J. Sci., (6): [5] Li Tao, ig-jug Seow ad Vijaya K. Asari, 6," Noliear Image Ehacemet to Improve Face ", Iteratioal Joural of Computatioal Itelligece Research.ISSN Vol., No.4, pp ,. [6] Osma N. ad Capt. Eder, 7," A oliear Techique for The ehacemet of extremely o-uiform lightig images ", joural of aeroautics ad space techologies V. 3 No.,. [7] Soy Wie S., J ad Hore R.E., 998 The Color Image Processig Had Book, Iteratioal Thomso,. 7