采用TV及纹理合成技术的分层图像修复  

Total variation and texture synthesis applied to multi-level image inpainting

作　　者：周姗姗[1] 赵知劲[1] 叶学义[1] 

机构地区：[1]杭州电子科技大学通信工程学院,杭州310037

出　　处：《计算机工程与应用》2011年第23期201-203,211,共4页Computer Engineering and Applications

基　　金：国家自然科学基金No.60802047;浙江省科技计划重点项目(No.2008C21092)~~

摘　　要：整体变分(TV)模型在图像修复中能够保持图像的边缘且数值实现方便,但在图像修复中对于人类视觉连通性的处理还有所不足。根据图像遗失或者损坏的不同类型,针对TV、CDD模型在图像修复中存在的问题,提出了一种结合TV、CDD模型及基于块的纹理合成算法的分层图像修复算法。实验结果表明,这种分层修复的方法在图像的结构修复和纹理修复两方面实现了较好的统一,而且在较大区域图像修复上表现出良好的效果。Although the Total Variation（TV） model is good at maintaining the edge of damaged images and reducing numerical calculation in image inpainting,it needs more improvement in the vision connectivity.A new image multi-level-inpainting method based on TV model and texture synthesis is proposed.The TV model is used to restore the shape information of missing area in an image.The inpainting result visual connectivity is improved by the CDD model.To the large texture area of the inpainted image,the patch based texture restoration model is used to process the obvious trace insulted from two previous levels inpainting.Experimental results show that the inpainting performance is better both in the structure and the texture.Moreover,the approach can be used in the image inpainting with bigger damaged area.

关 键 词：图像修复 整体变分(TV) 曲率驱动扩散(CDD)模型 分层修复算法 纹理合成 最小均方误差(MSE) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]