Study Of Perceptual Contrast Evaluation Methodology And Modeling Based On Color Appearance

Various types of display devices have been extensively applied in daily life and work with the development of science and technology and the improvement of the living standard. And the evaluations of their color characteristics have become the key to win in the hot competition among different display technologies. Hereinto, contrast is the representative parameter for the performance evaluation of displays. However, the traditional evaluation methods of contrast emphasize more on the physical measurement of luminance, which usually deviate from the real perception of the human visual system (HVS) and do not adapt to the variations of viewing conditions. Hereby, by performing the psychophysical experiments in this study, plenty of data were obtained for the visual threshold and perceptual contrast via simple neutral color patterns, so that the corresponding perceptual contrast models were built based on the color appearance model. Then the impact of spatial frequency and color on the perceptual contrast were analyzed through the visual measurements of the contrast sensitivity functions (CSF) of luminance and color grating patterns. Lastly, on the basis of the color appearance model and CSF, the perceptual contrast model of complex image was proposed, which was tested and verified using the visual data of image color appearance perception experiments.The color characteristics of displays were first measured and analyzed in detail before the visual experiments. For the large size liquid crystal displays (LCD) which do not fully meet the two assumptions of chromaticity constancy and channel additivity, the method of local three-dimension look-up table was adopted for their colorimetric characterization, and an auto-adaptation real-time searching algorithm was developed to promote the transformation efficiency with the characterization accuracy being better than0.5CIELAB color difference unit (△E*ab). For the cathode ray tube (CRT) display, the method of piecewise linear interpolation assuming variation in chromaticity coordinates (PLVC) combined with the dithering technique was implemented to achieve the high colorimetric characterization accuracy of0.31 △E*ab. The professional Eizo LCD. which approximately satisfies the two mentioned assumptions, was colorimetrically characterized using the method of piecewise linear interpolation assuming constant chromaticity coordinates (PLCC) to result in the characterization accuracy of0.98△E ab.The visual thresholds of19target stimuli with different lightness levels on the uniform neutral color background under9viewing conditions were measured using the psychophysical method of constant stimuli and data processing technique of probit analysis, which were represented by four different color attributes, i.e. luminance L in cd/m2, CIELAB lightness L*, CIECAM02lightness J and brightness Q. The prediction performances of different representation equations for visual threshold contrast were tested via the coefficient of variation (CV). which results indicated that the representation equations based on color appearance attributes were in better agreement with the visual data under different viewing conditions. This provided the experimental and theoretical supports for modeling the perceptual contrast.Based on the principle of the magnitude estimation method, two perceptual contrast assessment experiments were carried out, respectively, i.e. Experiment No.1with the reference and the test patterns under the same viewing condition and Experiment No.2with the reference pattern under the fixed viewing condition. Then two categories of perceptual contrast models were established through the fitting of visual data. One category is represented using the perceptual attributes especially the color appearance attributes, and another representation is based on the steps of the just noticeable difference (JND) of luminance. The prediction performance of these models were tested and compared by the ways of standardized residual sum of squares (STRESS) and F-test, and the results showed that the contrast model CQ adopting the CIECAM02brightness Q agreed better with the nonlinear perception characteristics of HVS than the other models and was well adaptive to different viewing conditions. Therefore, the CQ model may be suggested as the perceptual contrast model in practical applications to replace the traditional luminance ratio. In order to apply the CQ conveniently into the contrast evaluation of LCD. the spot test patterns were designed and the evaluation workflow involving the chromaticity prediction model of LCDs was proposed, of which the feasibility of the prediction procedure and the validity of the prediction performance were verified via the practical test of visual data.Utilizing the psychophysical method of interleaved staircase, the luminance/lightness and color contrast thresholds were measured, respectively, at gray and chromatic color centers with the spatial frequency ranged from0.5to11.7cpd on the CRT display. And the CSFs of different color centers and the chromaticity ellipses for color contrast thresholds at different spatial frequencies were obtained by fitting with the visual data. Then the influences of spatial frequency and chromacity on the contrast perception were analyzed. Meanwhile, the lightness and chromatic CSFs defined in the same scale by color difference were proposed, and the mathematic model of lightness CSF of different lightness levels at gray center was built from the Barten model.The complex images usually involve plenty of spatial and color information. Through the psychophysical method of category judgments, the image color appearance attributes were visually estimated, including image quality, contrast, colorfulness, sharpness, naturalness, global contrast and local contrast. On the top of the analysis of the image perceptual contrast and its influence factors, two image perceptual contrast models S-PBCDR-Ⅰ and S-PBCDR-Ⅱ were proposed, respectively. based on the CSF spatial adaptation and the color appearance model, and they were further compared with the relative single-image perceived contrast model△SIPk and the pixel-based color difference ratio model PBCDR. Through the test of visual data, the model S-PBCDR-Ⅰ exhibited best prediction performance in the perception of image contrast among all the models and was more simple and convenient as well in applications.Finally, the main contents and research achievements of this dissertation were summarized, and the perspectives of the future study were also forecasted.