Title:A propagation matting method based on the Local Sampling and KNN Classification with adaptive feature space

Abstract:Closed Form is a propagation based matting algorithm, functioning well on images with good propagation . The deficiency of the Closed Form method is that for complex areas with poor image propagation , such as hole areas or areas of long and narrow structures. The right results are usually hard to get. On these areas, if certain flags are provided, it can improve the effects of matting. In this paper, we design a matting algorithm by local sampling and the KNN classifier propagation based matting algorithm. First of all, build the corresponding features space according to the different components of image colors to reduce the influence of overlapping between the foreground and background, and to improve the classification accuracy of KNN classifier. Second, adaptively use local sampling or using local KNN classifier for processing based on the pros and cons of the sample performance of unknown image areas. Finally, based on different treatment methods for the unknown areas, we will use different weight for augmenting constraints to make the treatment more effective. In this paper, by combining qualitative observation and quantitative analysis, we will make evaluation of the experimental results through online standard set of evaluation tests. It shows that on images with good propagation , this method is as effective as the Closed Form method, while on images in complex regions, it can perform even better than Closed Form.