Title :
Simultaneous cartoon and texture for nonconvex image inpainting via the balanced approach
Author :
Yu-Lian Wu ; Xiang-Chu Feng ; Liang Luo
Author_Institution :
Sch. of Sci., Xidian Univ., Xi´an, China
Abstract :
This paper describes an image inpainting algorithm based on two tight frame systems that can sparsely represent cartoons and textures respectively. The proposed minimization formulation adopts nonconvex approximation via balanced approach, and an iterative algorithm is derived to find its solution. Numerical simulations demonstrate that the proposed nonconvex algorithm can significantly improve the image inpainting quality over the usual l1 algorithm.
Keywords :
approximation theory; concave programming; image reconstruction; image representation; image texture; iterative methods; minimisation; balanced approach; frame systems; image inpainting quality; iterative algorithm; l1 algorithm; minimization formulation; nonconvex approximation; nonconvex image inpainting algorithm; simultaneous cartoon representation; texture representation; Abstracts; Cartoons and texture; Image inpainting; Nonconvex; Tight frame;
Conference_Titel :
Wavelet Analysis and Pattern Recognition (ICWAPR), 2013 International Conference on
Conference_Location :
Tianjin
Print_ISBN :
978-1-4799-0415-0
DOI :
10.1109/ICWAPR.2013.6599329