The most two important properties of color spaces are without cross contamination between color channels （i. e. , lightness, saturation and hue） and with maximum perceptual uniformity. Based on multi-grid method and use standard color-difference formulas to construct color look-up table, realize transformation of images from initial color space to new color space. Validity of new color space is verified in image segmentation, using ISODATA algorithm to get the optimal initial cluster center and cluster number in k-means image segment. Experimental results show compared with other color spaces, each channel of new working space has high degree of independence and greatly meet human perception, which can improve precision image segmentation algorithms.