A Kernel Method with Manifold Regularization for Interactive Segmentation

Abstract

Interactive segmentation has been successfully applied to various applications such as image editing, computer vision, image identification. Most of existing methods require interaction for each single image segmentation, which costs too much labor interactions. To address this issue, we propose a kernel based semi-supervised learning framework with manifold regularization for interactive image segmentation in this paper. Specifically, by manifold regularization, our algorithm makes similar superpixel pair bearing the same label. Moreover, the learned classifier on one single image is directly used to similar images for segmentation. Extensive experimental results demonstrate the effectiveness of the proposed approach.