Because the K-means algorithm can't segment small targets in color images well,especially the images whose target color can't significantly distinguish from the background color. An improved color image segmentation algorithm based on sharpness (S-K-means) is proposed. The image sharpness evaluation theory is introduced and gray matrixes of R,G,B color channels are transformed into three sharpness matrixes,then they are converted into Lab color space. At last blocks are classified by K-mean clustering method. The new algorithm is applied to segment a number of power line images and compared with other segmentation methods by experiments in order to verify the effectiveness of the proposed algorithm. At the end,both theoretical analysis and experimental results are given. Experimental results show that the effect of this algorithm is better,the conclusion can provide a reference for high voltage cable patrol in electric power system,and this algorithm has high availability.