摘要
提出一种基于L1-TV模型的图像修复方法。在正则化框架下采用L1范数度量逼近项,增加了模型修复图像破损部分的能力。利用"变量分裂"思想和半二次光滑化技术,克服L1-TV模型的算法求解困难,使得新模型的解易于求出。与两个已有算法的对比实验结果表明,该算法在获得较高的修复效果的同时,可以显著地降低运算时间。
An image inpainting algorithm based on L1-TV model is introduced. The utilization of L1 norm in the fidelity term in- creases the capacity of restoring the damaged parts of images. The new model uses variable splitting and half-quadratic techniques to o- vercome the algorithmic difficulties for the numerical treatment of L1-TV model. Compared with the two existing algorithms, the experi- mental results show that our algorithm can excellently recover the information in the damaged parts of images, as well as dramatically reduce the computational time.
出处
《济南大学学报（自然科学版）》
CAS
北大核心
2011年第4期371-376,共6页
Journal of University of Jinan(Science and Technology)
基金
国家自然科学基金(60975083)
