The Effects of Skin Tone Posterization in Image Reproduction Workflow

• Main Authors: Yu-chuan Chen, 陳郁娟
• Other Authors: Pei-li Sun
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/q5gt36

碩士 === 世新大學 === 圖文傳播暨數位出版學研究所(含碩專班) === 96 === Posterization is an unpleasant color discontinuity phenomenon in image reproduction. It is frequently happened in hardcopies but difficult to be observed during their pre-press stage. It would result in delayed delivery and lose considerable money in reprint the reproductions. To solve this problem, the study aims to explore the source of skin tone posterization in image reproduction workflow. The findings were used to design a model to show ‘Posterization Index’ of an original for predicting the final print posterization in the pre-press stage. The study chose typical skin spectral reflectances of four different races from ISO/TR16066-2003 report to generate color gradations. Four experiments were performed. The first one, compared sRGB values of the skin gradations with their actual color measurements on a high-end and a low-end LCDs under dark and bright viewing environment. The results show that the low-end LCD is not reliable and bright environment compresses dark color steps significantly so as to reduce the possibility to inspect the posterization effect visually on dark regions. The second experiment compared posterization effects of two printers under two reproduction modes. The results show both printers compress dark-tone heavily and overstate their chroma in their mid-tone during computation. The third experiment asked observers to judge visual thresholds of posterization on a display and a printer using six series of skin tones. The results show that the thresholds among dark/middle/bright regions are device dependent. It also shows that Mach band effect is weakened by blurring. The fourth experiment asked observers to identify posterization areas on portrait photos with 4 different lighting directions. The results show that the lighting directions have something to do with the effect. Finally, we derived a model based on Laplacian of Gaussian filtering with the above experimental data to process the original photos to predict the degree of posterization on their final prints. The proposed ‘Posterization Index’ has good correlation (coefficient about 0.6) to that of visual posterization.