【作者】 张静； 余烨； 叶强强； 刘晓平；

【Author】 Zhang Jing Yu Ye Ye Qiangqiang Liu Xiaoping School of Computer & Information,Hefei University of Technology,Hefei,230009

【机构】 合肥工业大学计算机与信息学院；

【摘要】 图像修复是利用一些算法对产生划痕和有缺损的图像进行修复,或者从图像中去除指定的物体和文字,以达到特定的目的。本文提出了一种基于片相似性的BSCB模型,利用片相似性各向异性扩散的性能来对受损区域进行平滑降噪,并针对结构简单的人工图像和有复杂结构的自然图像进行了修补。实验结果表明,基于片相似的BSCB模型可以获得良好的修复效果,且与BSCB-ASG模型和TV模型相比,可以减少图像的修补次数,改善图像的修补效果,因而可以满足图像修复的需要。更多还原

【Abstract】 Image inpainting is used to restore the damaged images which have cracks and scratches or to remove the specified objects and texts in order to reach specific goals.This paper proposed an improved BSCB model based on patch-similarity,which exploits property of patch-similarity anisotropic diffusion for denoising of smooth with damaged image areas,meanwhile inpaints the manual image of simple structure and natural image of complicated structure.Experimental results show that the improved BSCB model can achieve good result of inpainting.Comparing with the BSCB-ASG model and the TV model,the BSCB model based on patch-similarity can reduce inpainting time and improve the inpainting effect of image,therefore it can fulfill the demands of image inpainting.更多还原