Illumination-estimation On Chromaticity Based On Specular Reflection Area Of Images

In the color management,it often affects from the surrounding illumination to achieve the image-reproducing in different environments on devices.Thus,it is necessary to estimate the illumation of device.The thesis treats achieving color constancy as the starting-point and destination-point of the Illumininant-estimation research about images,and according to summarize the exsiting algorithms,especially the algorithm based underlying characteristics of images,grasps the characteristics of formation of specular reflection pixels in images including specular reflection area,about the problem of separation specular reflection pixels,the thesis improves or raises two approachs: adaptive region-segmentation,image-segmentation based on gradient,from two different aspects including precision and adaptability approach;and also,after separating specular reflection pixels from the image,it improves neural network method,raises polynomial-fitting method and the Center-Probability method based on Inverse-Intensity Chromaticity Space in order to estimate the illumination chromaticity,For the three methods,dichromatic reflection model is the basis of study.The result of experiment indicates that: the proposed algorithms are better than the traditional algorithm in precision and adaptation.