Fast algorithms for color image processing by principal component analysis

This paper discusses a new approach for ordering color vectors by principal component analysis. A color image is represented by a vector field and the color vectors in an n by n pixel window are ordered according to the projection scores obtained by projecting each color vector within the window on the principal axis. We subtract each color vector in a window from the color vector of the central pixel before constructing the corresponding covariance matrix. For the purpose of computation efficiency, a fast approximation of the principal axis is also proposed. By applying the vector order statistics, various applications of color image processing, such as color image sharpening, color image compression, and color edge detection are also proposed in this paper. Therefore, the proposed color vector ordering method can be,used as a tool for color image processing. (C) 2003 Elsevier Science (USA). All rights reserved.