Learning-Based Weighted Total Variation for Structure Preserving Texture Removal

摘要：

　　An image is generally formed as the composition of salient structures and complex textures.While structures are important for human perception and image analysis,structure extraction from textures remains a challenging issue to be investigated.Even though several methods have been proposed to do this job,they commonly have to balance between texture removing and structure preservation.One problem is that few methods take structural contours into consideration.In this paper,we propose a new learning-based weighted total variation(LTV)model,where the weights are learned from different kinds of texture images to well discriminate pixels belonging to structural contours from pixels belonging to textures.The Chambolles projection method is utilized to solve the optimization problem.Experimental results show that compared with the competing methods,the proposed algorithm performs better in preserving sharp structures while removing textures.

关键词： Structure extraction Contour detection Learning-based total variation Chambolles projection

作者: Shoufeng Zheng Chunwei Song Hongzhi Zhang Hongzhi Zhang Wangmeng Zuo

作者单位: School of Computer Science and Technology,Computational Perception and Cognition Center,Harbin Institute of Technology,Harbin 150001,China

会议类型: 国际会议

会议名称: 第七届全国模式识别学术会议(The 7th Chinese Conference on Pattern Recognition,CCPR2016)

会议地点: 成都

会议语种:英文

页码: 147-160

在线出版日期: 2016-11-03（万方平台首次上网日期，不代表论文的发表时间）