Interactive Clothing Segmentation Based on Jseg and Grabcut

Clothing segmentation is a challenging issue of research, which is gaining more and more attention. This paper presents an interactive clothing segmentation method based on JSEG and GrabCut. Firstly, we pre-process the image from pixels to regions using JSEG algorithm for the purposes of lessening computer loads and reducing the noise of image, and then build a graph using regions as nodes, and define the seeds of clothing or non-clothing areas interactively. Finally, solve the energy function of the graph using min-cut (max-flow) approach, and separate the clothing from the image. In some case of inaccurate boundary, we adopt repeated segmentation technology to repair the results. Some given samples show that the approach is effective and can achieve promising results.