AN ASSESSMENT OF COLOR DIFFERENCE METRICS BASE ON ROC CURVE

• Main Authors: Tsung-yen Yeh, 葉宗諺
• Other Authors: Ray-chin Wu
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/2z2kem

碩士 === 大同大學 === 光電工程研究所 === 102 === According the colorimetric has been developed so far, it is better and better about presented color. However, which color expression of the color spaces would be the most outstanding and the most in line with the human eye perceives for image? In this study, Subjects examined image pairs consisting of an original and a reproduction on LCD monitor. We can get data about viewing digital color of human visual system (HVS) and refer to the distribution of error marks by the subjects as image distortion maps. The imaging distortion maps were obtained from the statistical results of the subjects’ perceptual markers. The color difference pixel by pixel between original and reproduced images for various metric RGB RMSE, CIELAB, S-CIELABΔE, CIEDE2000ΔE00, CIECAM02, ICAM06 were calculated. Based various threshold value of color difference metrics and corresponding image distortion maps, the hit rate (HR) and false alarm rate(FAR) were obtained to plot the receiver operating characteristic (ROC) curve. According the area under curve (AUC), the validity of the color difference metrics was evaluated. As the results, it is found the validity of the color difference metric for digital image is depend on the textual content. Both S-CIELAB and iCIECAM06 predictions incorporating spatial pre-processing step had better agreement with the perceptual data.