Image super pixel segmentation approach using the lazy random walk (LRW) algorithm with self-loops has the qualities of segmenting the pathetic restrictions and convoluted texture regions as extremely glowing by the new global probability maps and the transform instance policy. Our technique begins with initializing the seed positions and runs the LRW algorithm on the input image to gain the probabilities of each pixel. Then boundaries of original super pixels are obtained according to the probability and the commute time. Theoriginal super pixels are iteratively optimized by the energy utility, which is defined on the commute time and the texture quantity.The performance of super pixel is improved by relocating the midpoint positions of super pixels and isolating the large super pixels into miniature ones with optimization algorithm. The experimental results have confirmed that our method achieves recoveredact than preceding super pixel approaches.