Upcoming SlideShare
Loading in...5

Like this? Share it with your network



Uploaded on


  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Be the first to comment
    Be the first to like this
No Downloads


Total Views
On Slideshare
From Embeds
Number of Embeds



Embeds 0

No embeds

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

    No notes for slide


  • 1. M.Sangeetha / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 3, Issue 1, January -February 2013, pp.1532-1536 Analysing Image Quality via Color Spaces M.Sangeetha1 1 Department of Computer Science , Sri Sarada Niketan College for Women, Amaravathi Pudur, Karaikudi- 630301, Tamilnadu, India.Abstract Color space is the mathematical axis spaces: It results from other spaces by applyingrepresentation of a set of colors. A large amount mathematical operations that aim at decor relatingof color spaces have been proposed for different individual components.purposes. But the RGB color space is a basic andwidely used color space, and other color spaces This work proposes that, RGB color spaceare usually defined by transformation of the are transformed into different color spaces andRGB color space. This paper focuses on the applied in images. Next, separate images have to beapplication of Other color spaces in images by analyzed, to find the certain image quality factors.transforming the RGB color space. Also a Image qualities should be better with another thatstatistical analysis of the sensitivity and color space is good. The impact of this proposedconsistency behavior of objective image quality work is all are known RGB is the first and best colormeasures are performed. So that resulting the space. But what is the next color space we are notimage qualities can be decided that which Color known. In this proposed work to analyze whichspace is better without the interference of RGB color space better and provide rank of color spacecolor space. depends upon the quality factors. This work is an extension of Jian Yang, ChengjunLiu, and LeiZhangKeywords: Color Space, Image Quality, RGB which used Color space normalization techniquescolor space. used in face recognition. Here, we introduce the new method for color space apply in whole image and1. Introduction detect the quality of the image. In the general way, Color is the brains reaction to a specific the simplicity of implementation, the low processingoptical input signals from the eye. The retinas in our cost and the high quality of results can beeyes though have three types of color receptors in considered as the main contributions of our work.the form of cones. We can actually only detect three Also, we verified that our methods are more robustof these visible colors - red - blue and green. These than other implemented methods.colors are called primary color space. In this three This paper is organized as follows. Incolors that are mixed in our brain to create several Section 2 we present a related works. In Section 3different “sensations” of the color. These sensations we propose a color space conversion. In Section 4have been defined by the CIE different Measuring we propose a quality measures to find the imageColor like Brightness, Hue, Colorfulness, Lightness, quality. In section 5 gives my proposed works inChroma, and Saturation. So this measuring color this paper. In Section 6 we report on a experimentalcombination to produce the different color space analysis for both color space conversion result andfrom primary color space. comparative analysis for all color spaces. Some conclusions and a summary of future works are A color space is a method of representing given in Section 6.each color in terms of a combination of severalnumeric values. It is a method by which we can 2. Related Works:specify, create and visualize color. Different color Jian Yang, ChengjunLiu, LeiZhang paperspaces are better for different concerned presents the concept of color space normalizationapplications. Now days we can use several color (CSN) and two CSN techniques for enhancing thespace. Some of the color spaces are,RGB, discriminating power of color spaces for faceCMY,XYZ,xyY,|1|2|3|,UVW,LSLM,L*a*b,L*u*v, recognition. The experimental results reveal thatLHC,LHS,HSV,HSI,YUV,YIQ,YCbCr. All these some color spaces, like RGB and XYZ, arecolor spaces under the category of primary relatively weak for recognition, where as other colorspaces: It is based on the trichromatic theory which spaces, such as I1I2I3, YUV, YIQ and LSLM, arestates that any color can be expressed as a mixture relatively powerful.of three primaries. Luminance-chrominance: It İsmail Avcıbas, Bülent Sankur statisticalrepresents colors in terms of luminosity and two analysis of the sensitivity and consistency behaviorchromaticity components. Perceptual spaces: It is of objective image quality measures. We categorizebased on the human visual system and commonly the quality measures and compare them for stillstrives for being perceptual uniform. Independent image any applications. The measures have been 1532 | P a g e
  • 2. M.Sangeetha / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 3, Issue 1, January -February 2013, pp.1532-1536categorized into pixel difference-based, correlation- compute chroma, by multiplying saturation by thebased, edge-based, spectral-based, context-based maximum chroma for a given lightness or value.and HVS-based (Human Visual System-based) Next, we find the point on one of the bottom threemeasures. faces of the RGB cube which has the same hue and chroma as our color (and therefore projects onto the3. Color space conversion same point in the chromaticity plane). Finally, we Color space conversion is the translation of add equal amounts of R, G, and B to reach thethe representation of a color from one basis to proper lightness or value.another. This typically occurs in the context ofconverting an image that is represented in one colorspace to another color space, the goal being to makethe translated image look as similar as possible tothe original. All the image get color informationfrom which contains only RGB values. In the RGBcolor space each color appears as a combination ofred, green, and blue. RGB is a basic color space forcomputer graphics because color displays use red, 3.4. RGB to YCbCr conversiongreen, and blue to create the desired color. This It is used as a part of the color imagemodel is called additive, and the colors are called pipeline in video and digital photography systems.primary colors. Basically the color space value of Y is the luma component and CB and CR are theRGB is to be known for formulating the other color blue-difference and red-difference chromaspace conversion. In this paper we concentrate components. The equations of the RGB to YCbCrfollowing color space conversion. are formed in a way that rotates the entire nominal RGB color cube and scales it to fit within a (larger)3.1. RGB to YIQ conversion YCbCr color cube. The YIQ color space was formerly used inthe National Tele-vision System Committee (NTSC)television standard. The Y component represents theluma information, and is the only component usedby black-and-white television receivers. I and Qrepresent the chrominance information. The YIQ 3.5.RGB to L*a*bsystem, which is intended to take advantage of It is more often used as an informal abbreviation forhuman color-response characteristics, and can be the CIE 1976 (L*, a*, b*) color space , whosederived from the corresponding RGB space as coordinates are actually (L*, a*, and b*). The colorfollows. spaces are related in purpose, but differ in implementation. Perceptually uniform means that a The L*a*b* color space includes all perceivable colors which means that its gamut exceeds those of the RGB and CMYK color models. One of the most important attributes of the L*a*b*-model is the device independency. This means that the colors are3.2. RGB to XYZ conversion defined independent of their nature of creation. The XYZ color space was derived from aseries of experiments in the study of the humanperception by the International Commission onIllumination.The transformation from the RGBcolor space to the XYZ color space is as follow3.3. RGB to HSI conversion 4. Quality Measures HSI means Hue ,Saturation ad Intensity for Conversion of an image in each color spacehuman perception. It represents color s similarly the quality may be differing from one color space tohow the human eye senses colors.The conversion other depending on the color space. In the case ofstarts form RGB to HIS as follows. First, we conversion image, it is important to reproduce the 1533 | P a g e
  • 3. M.Sangeetha / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 3, Issue 1, January -February 2013, pp.1532-1536image close to the original image so that even thesmallest details are readable. Conventional measuresare designed to quantify the error, sensitivity 4.3 Average Differencebetween the original image and the converted one, A lower value of Average Difference (AD)while keeping most of the signal characteristics gives a “cleaner” image as more noise is reducedintact. Imaging systems may introduce some and it is computed Using Table(1)amounts of distortion or artifacts in the signal, so thequality measures are an important problem. There 4.4 Maximum Differenceare several techniques and metrics that can be Maximum difference (MD) is calculatedmeasured objectively and automatically evaluated using Table(1) and it has a good correlation withby a computer program. Therefore, they can be MOS for all tested conversion images so this isclassified as full-reference (FR) methods and no- preferred as a very simple measure as a reference forreference (NR) methods. In FR image quality measuring conversion image quality in differentassessment methods, the quality of a test image is color spaces. Large value of MD means that theevaluated by comparing it with a reference image image is of poor quality.that is assumed to have perfect quality. NR metricstry to assess the quality of an image without any 4.5 Normalized Correlationreference to the original one. In this paper we are The closeness between two digital imagesusing FR methods to measure the quality from can also be quantified in terms of correlationoriginal RGB color space Image to Converts Other function. These measures measure the similaritycolor spaces images. between two images like a original color space in the image other one converted color space image,4.1 Mean Square Error hence in this sense they are complementary to the In the image processing the most difference based measures. All the correlation basedfrequently used measures are deviations between the measures tend to 1, as the difference between twooriginal and reconstructed images of which the images tend to zero. As difference measure andmean square error (MSE) or signal to noise ratio correlation measures complement each other,(SNR) being the most common measures. The minimizing Distance measures are maximizingreasons for these metrics widespread popularity are correlation measure and Normalised Correlation istheir mathematical tractability and the fact that it is given Table(1).often straightforward to design systems thatminimize the MSE but cannot capture the artifacts 4.6 Normalized Absolute Errorlike blur or blocking artifacts. The effectiveness of Normalized absolute error computed bythe coder is optimized by having the minimum MSE table(1) is a measure of how far is the conversionat a particular conversion. Calculated MSE in color image from the original image with the value of zerospace image into original image to use the Table(1). being the perfect fit . Large value of NAE indicates poor quality of the image.4.2 Peak signal-to-Noise Ratio Larger SNR and PSNR indicate a smaller 4.7 Structural Contentdifference between the original (without noise) and Correlation, a familiar concept in imagereconstructed image. The main advantage of this processing, estimates the similarity of the structuremeasure is ease of computation but it does not of two signals. This measure effectively comparesreflect perceptual quality. An important property of the total weight of an original signal to that of aPSNR is that a slight spatial shift of an image can coded or given. It is therefore a global metric;cause a large numerical distortion but no visual localized distortions are missed .This measure isdistortion and conversely a small average distortion also called as structural content. The Structuralcan result in a damaging visual artifact, if all the content is given by Table(1) and if it is spread at 1,error is concentrated in a small important region. then the converted image is of better quality andThis metric neglects global and composite errors large value of SC means that the image is poorPSNR is calculated using Table(1). quality . 1534 | P a g e
  • 4. M.Sangeetha / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 3, Issue 1, January -February 2013, pp.1532-15365. Methodology quality wise image for each color space ,the high Hundred images, twenty images each in quality of the image should be identified. ThatRGB to YIQ color space, RGB to YCbCr color image color space better for other color spaces. Inspace , RGB to HSI color space , RGB to L*a*b this proposed techniques we are conversion fromcolor space and RGB to XYZ color space, of size five color spaces each color space calculatedconverted each image equal size for the analysis. quality measure.The conversion ratio and other quality measuresare computed using the original and coded images. 6. Result and DiscussionIn an our analysis original image is RGB color 6.1 Experimental Results:space image and converted other color space In this section, we show the result forimage. If the quality of the image should be RGB color space into different color spacescalculated above quality measure for each color applying for original along with the image. So we obtain the Figure 1: Original RGB color space Image RGB to HSI RGB to L*a*b RGB to YIQ RGB to YCbCr RGB to XYZ Figure 2: Conversion of other color space Images 1535 | P a g e
  • 5. M.Sangeetha / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 3, Issue 1, January -February 2013, pp.1532-15366.2.Comparative Analysis Result: Enhancing the discriminating power ofTable computed values of the Conversion color spaces for face recognition”.different color space values and other objective [4] I. Avcibas, B. Sankur, K. Sayood,quality measures obtained. “Statistical Evaluation of QualityConversion from RGB color space into Measures In Image Compression”, J. of Electronic Imaging, (in review).7. Discussions and Conclusions: [5] Adrian Ford and Alan Roberts August 11, The conversion ratio for RGB to YIQ 1998(b),” Colour Space Conversions”.color space, RGB to YCbCr color space, RGB to [6] Bauer S., Zovko-Cihlar B. and Grgic M.HSI color space, RGB to L*a*b color space and (1996a) „Objective and SubjectiveRGB to XYZ color space obtained MSE and PSNR Evaluations of Picture Quality in Digitalimage quality factor is better in RGB to L*a*b Video Systems‟,Multimedia Technologycolor space the result is 7308.84 for MSE and 9.49 and Digital Communication Services,PSNR with low value indicates a good quality lCOMT 1996, pp. 145-150.image and that is observed with the value. Average [7] Hamid Rahim Sheikh, Muhammad Farooqdifference of the each color space is varying from Sabir and Alan C. Bovik (2006) „Aone to other color spaces. The RGB to L*a*b color statistical evaluation of recent fullspace obtained value is very low for 5.02 so the reference image qualityresult gives a “cleaner” image as more noise is assessmentalgorithms‟, IEEE Trans Imagereduced and better quality of the image. Processing, Vol. 15, No. 11, pp. 3440- The Structural content result smaller value 3451.of SC means that the image is good quality . In an [8] Richard Dosselmann and Xue Dong Yangour result obtained SC value is 1.00 in RGB to (2005) „Existing and emerging imageL*a*b color space is smaller value. So in SC quality metrics‟, CCECE/CCGEI, pp.depends upon the Image quality better color space RGB to L*a*b.Large value of Normalized [9] Sonja Grgic, Mislav Grgic and BrankaAbsolute Error indicates poor quality of the image. Zovko-Cihlar (1996) „Objective andIn an our result obtained NAE value is 0.17 in RGB subjective measuruments for videoto L*a*b color space is smaller value so better compression system‟,Proceedings of thequality. Also in NAE depends upon the Image International Conference of Multimediaquality better color space is RGB to L*a*b.So our Technologies, ICOMT 1996, pp. 145-150.main contribution result is overall image applied [10] Tomas Kratochvil and Pavel Simicekany color spaces is better quality obtained for RGB (2005) „Utilization of MATLAB forto L*a*b. picture quality evaluation‟, Institute of From this work, we observed that the Radio Electronics, Brno University ofcolor space conversion present an perfect image Technology, Czech. Republic.quality for each color space and basis for futuredevelopments such as: (i) we are implemented only5 color space from RGB color spaces; but now adays there are several color spaces used .so convertall the color space from the RGB color space that ismore effective (ii) we are implemented only somequality measure to find the quality of the image, so SangeethaMuthaiah received herimplemented different quality factor to be used in B.Sc Degree from Madurai KamarajUniversity,future enhancement and obtain the better quality Madurai. She also received M.Sc.,and M.Phil.,result. Degree from Alagapppa University, Karaikudi inReferences: 2008, 2010 respectively. She is currently working [1] Rafael C. Gonzales. Digital Image as Lecturer, Department of Computer Science, Sri Processing. Tom Robbins, 2 edition, 2002. [2] Users Guide Statistics Toolbox 7. Sarada Niketan College for Women, Amaravathi MathWorks Matlab. MathWorks Matlab, Pudur, Karaikudi-630301, Tamilnadu, India . Her 2010. research interest includes Data mining and Image [3] Jian Yang , ChengjunLiu , LeiZhang Processing ad hoc wireless Networks &Security ,2009,” Color space normalization: and Computer Algorithms. 1536 | P a g e