Color Constancy Based on Automatic Adjustment of Image Luminance and White Region Detection

• Main Authors: Hung-kai Hsueh, 薛鴻凱
• Other Authors: Cheng-ho Hsin
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/04958605945705686574

碩士 === 逢甲大學 === 通訊工程所 === 100 === Color constancy refers to the invariability of color perception for the same surface to human eyes under different illumination conditions. However, image capture devices have no similar mechanism. Current machine color constancy schemes are still unable to remove all kinds of color casts. The purpose of this thesis is to develop a novel color constancy method based on white region detection. Firstly, the luminance of an input image is automatically adjusted by the global and local lightness adaptation scheme. This step produces an output image with the most appropriate brightness level and increases the visibility of the original image details. Then a recursive algorithm based on a sequence of color saturation thresholds is applied to search for possible white regions in the output image. Once a true white region is identified, it can be used to estimate the color of the illumination. The input image is corrected as under a canonical illumination, hence the goal of the color constancy is achieved. The proposed color constancy scheme was tested for a variety of image databases. The experimental results were also compared to that of other state-of-the-art methods. The proposed method consistently performed well. Keyword：color constancy, white region detection, illumination estimation.