| Summary: | 碩士 === 逢甲大學 === 通訊工程學系 === 102 === Most color constancy algorithms assume a single light source with uniform illumination across a given image but this assumption seldom holds in reality. The purpose of this thesis is to develop a white balance algorithm that can cope with two illuminant light sources. We have established two types of image luminance adjustment schemes for our white balance algorithm. The gray component of an image is adjusted in the first type scheme, and then the color of the original image is restored. An anchoring method that compensating red (R), green (G), and blue (B) channels is applied to eliminate mild color casting. On the other hand, the luminance of each R, G, B channel is adjusted independently in the second type scheme. This scheme can eliminate severe color casting which existing in an image with a proper anchoring process. The first type adjustment scheme is unable to correct severe color casting. The second type adjustment scheme fails when an image contains a dominant object color. We have derived the automatic criteria to judge the validity of these two types of luminance adjustment schemes. Since each valid type of luminance adjustment scheme gives a white-balanced image, we have developed a method to generate a pixel-wise illuminant color map which is applied to generate a final integrated white-balanced image.
|
|---|
