Morphology Algorithm using Conspicuousness Map for Color Image Processing

Abstract

In this paper, we propose a concept of Conspicuousness Map, and use it in mathematical morphology for color image processing. Unlike the previous technologies, we use quantized conspicuousness degree of pixel as the foundation of morphological processing rather than using the simple Euclidean distance or other similar method. Traditional methods which directly using pixel value as the foundation of morphology processing should lead to an easily neglected problem -- why the white is bigger than the black, or other similar relationship. As a result of this question, our algorithm don't care which color is bigger or smaller, but pay attention to which pixel is conspicuous. The results of image denoise experiment show a high performance in color image processing.