REDUCTION OF COLOR SPACE DIMENSIONALITY BY MOMENT-PRESERVING THRESHOLDING AND ITS APPLICATION FOR EDGE-DETECTION IN COLOR IMAGES

A method for reduction of color space dimensionality by moment-preserv ing thresholding and its application in edge detection for color image s is proposed. An input color image is partitioned into n x n non-over lapping blocks. A moment-preserving thresholding technique is then app lied individually to each color plane of each image block. Two sets of (R, G, B) tristimulus values are obtained from the thresholding resul ts to form two representative color vectors for each block. The differ ence vector between these two representative color vectors is used as an axis onto which all the data in the block are projected to reduce t he color space to one dimension, A single-spectral image block is so o btained. Due to the use of analytic formulas in the thresholding step, the proposed dimensionality reduction method is found faster than the KL expansion or vector median approaches which are also applicable fo r dimensionality reduction. An (n + 1)x(n + 1) circular window is sele cted to sample the resulting single-spectral image, which in turn incl udes the n x n square block. Some mass moments of the window data are computed and used for edge detection in the circular window. Due to th e use of the larger detection window which results in smaller overlapp ing detection areas, the computation time for the edge detection step is reduced, compared with other similar approaches using overlapping d etection windows. Experimental results show that the proposed approach is effective in reduction of color space dimensionality and edge dete ction in color images.