A new method of reducing the dimensions of color vector (CV) representation is proposed. By analyzing the color data distribution of true color digital micrographs (TCDM) of human leukocyte in I 1I 2I 3 CV representation spaces, the 3D CV representation is reduced to 2D I 2I 3 CV representation based on the human visual system. Employing the I 2I 3 CV representation as the CV representation of images, the TCDMs of human leukocyte is segmented with back-propagation (BP) network. The experimental results show that, comparing with the I 1I 2 CV representation, by using this I 2I 3 CV representation method the CV representation is optimized, the convergence of BP network is accelerated, and the accuracy of images segmentation is increased.