Combining Co-Training and Co-Testing for Interactive Image Segmentation
Liantao Wang, Xuelei Hu, Jianfeng Lu · 2012
This paper presents an elastic framework with human interactions for image segmentation. Under this framework, we extract color and texture features for each pixel and consider them as two independent views. Based on pixel classification, the image segmentation can be accomplished through combining multi-view semi-supervised and active learning with human interactions. Experiments on various color images show the feasibility and effectiveness of our framework.