Color image morphology based on distances in the HSI color space

Multivariate mathematical morphology is increasing becoming a powerful utility for color image processing and analysis. In the paper, a new vector ordering scheme based on distances in the HSI color space is proposed. The methodology first uses nonlinear weighted combination with color distances, where vectors are reduced to scalars with the help of weighted coefficients. These coefficients are entirely determined by the saturation-intensity plane, which fit in with human perception. Then a lexicographical cascade is employed to complete total ordering. On the basis of the ordering approach, new erosion and dilation operations are defined. The performance of proposed is illustrated by means of color edge detection. Experimental results demonstrate its advantageous effectiveness. Additionally, the proposed framework can be also extended for other perceptual color representations based on polar coordinate.