SCID : Full Reference Spatial Color Image Quality Metric

The most used full reference image quality assessments are error-based methods. Thus, these measures are performed by pixel based difference metrics like Delta E ( E), MSE, PSNR, etc. Therefore, a local fidelity of the color is defined. However, these metrics does not correlate well with the perceived image quality. Indeed, they omit the properties of the HVS. Thus, they cannot be a reliable predictor of the perceived visual quality. All this metrics compute the differences pixel to pixel. Therefore, a local fidelity of the color is defined. However, the human visual system is rather sensitive to a global quality. In this paper, we present a novel full reference color metric that is based on characteristics of the human visual system by considering the notion of adjacency. This metric called SCID for Spatial Color Image Difference, is more perceptually correlated than other color differences such as Delta E. The suggested full reference metric is generic and independent of image distortion type. It can be used in different application such as: compression, restoration, etc. INTRODUCTION Objective approaches for image quality evaluation can be divided in three categories: full reference metrics (FR), reduced reference metrics (RR) and no reference (NR) metrics. Full reference metric approaches, requires besides the image at test a reference image, i.e a non distorted version of the same, to compare to. In the no-reference approach image quality is assessed using only the information content of the test image. The distortion estimation is performed only from the distorted version. In general they have lower complexity but are less accurate than FR metrics and are designed for a limited set of distortions and video formats. In the reduce-reference approach, only partial information about the original image is available. An RR metric defines what information has to be extracted form the original image, so it can be compared with the one extracted in the distorted version. In the literature, the most used FR image quality assessments are error-based methods [15]. Thus, these measures are performed by pixel based difference metrics like Delta E ( E), MSE, PSNR, etc. Therefore, a local fidelity of the color is defined. However, these metrics does not correlate well with the perceived image quality. Indeed, they omit the properties of the HVS. Thus, they cannot be a reliable predictor of the perceived visual quality. All this metrics compute the differences pixel to pixel. Therefore, a local fidelity of the color is defined. However, the human visual system is rather sensitive to a global quality. In this paper, we present a novel full reference color metric that is based on characteristics of the human visual system by considering the notion of adjacency. This metric called SCID for Spatial Color Image Difference, is more perceptually correlated than other color differences such as Delta E. The suggested full reference metric is generic and independent of image distortion type. It can be used in different application such as: compression, restoration, etc. In the case where no reference is available, this distance can be combined with the automatic perceptual image enhancement technique ACE [1] (Automatic Color Equalization). In this case the image at test is corrected with ACE, then the SCID distance is computed between the image at test and its ACE corrected version. Image Quality and System Performance VI, edited by Susan P. Farnand, Frans Gaykema, Proc. of SPIE-IS&T Electronic Imaging, SPIE Vol. 7242, 72420U · © 2009 SPIE-IS&T CCC code: 0277-786X/09/$18 · doi: 10.1117/12.806031 SPIE-IS&T/ Vol. 7242 72420U-1 In the next section, image quality assessment methods are described briefly, in particular the color difference metrics are introduced and discussed before introducing in section 3 our new spatial color metric. Experimental results and evaluating tests are presented in section 4. 1. IMAGE COLOR QUALITY ASSESSMENT Considerable work has been accomplished in terms of color difference perception. In 1942, MacAdam wanted to test the linearity of errors in the color matching experiment. He discovered that color differences are not linear. The most widely used color difference equations in the last decades are CIELAB and CIELUV color difference equations recommended in 1976 by the CIE. In both CIELAB and CIELUV color spaces, the color difference E* between two arbitrary colors is defined as an Euclidian distance in a uniform space comprising a lightness L* axis and red-green, yellow-blue component color axes using rectangular coordinates. The color differences in CIELAB color space are given by equation (1): (1) Where a* and b* are the redness-greenness and yellowness-blueness scales in CIELAB color space. The color difference formula CIE 76 is in many cases not adapted to human perception. The work in the area of color differences has concentrated on collecting reliable data and developing equations that describe the perceived color-difference results. Newer equations have been developed on base of the CIELAB (CIELCH) color space with application weighting difference components such as L*, C* and H*. Weighting functions SL, SC, SH are computed from regression analysis by using linear (CIE1994) or hyperbolic model. During the development of new color-difference formulas (CIE1994, C94CHR and MV-1(l:c)) there was considerable discussion about possible hue dependencies, as exemplified by the CMC [12] and BFD [2] equations. The CMC (l:c) color-difference formula was a refinement of the JPC79 equation developed by McDonald (1980). McDonald found that, for brown and purple-blue colors, CIELAB tolerances were over predicted. Therefore hue-angle dependent correction was implemented in CMC equation. The BFD colordifference formula was based on the Luo-Rigg (BFD) dataset. Luo-Rigg found that green colors were also over predicted. Bern’s studies on RIT and Du Pont dataset [13] showed that this hue dependency is not necessary condition by development of a new color-difference formula. Based on this study the CIE1994 color-difference equation was adopted, with the general form (2):