Interactive Clothing Image Segmentation Based on Superpixels and Graph Cuts

Abstract

In this paper, we propose an interactive clothing image segmentation method based on super pixels and Graph Cuts. Firstly, we process the image from pixels to super pixels with the method of SLIC to reduce the computational loads and lower the effect of noise, and then a graph is constructed using super pixels as nodes. Finally, min-cut/max-flow algorithm is applied to solve the energy function. In this process, the seeds of clothing or non-clothing are decided interactively. Some given samples show that the method can achieve promising clothing segmentation results.