Multi-class Video Objects Segmentation Based on Conditional Random Fields

Abstract: Video object segmentation has been widely used in many fields. A conditional random fields (CRF) model is proposed to achieve accurate multi-class segmentation of video objects in the complex environment. By using CRF, the color, texture, motion characteristics and neighborhood relations of objects are modeled to construct the corresponding energy functions in both the temporal and spatial domains. The model is trained with annotated samples by using LogitBoost classifier. The energy function is amended by adding a constraint factor which is used to indicate the interaction between two adjacent images in the video sequence. Experimental results show that the proposed algorithm can achieve high performance for multi-class objects segmentation in videos under complex environment. It can also get good recognition results when dealing with multi-viewed objects or serious sheltered objects. Copyright © 2014 IFSA Publishing, S. L