基于稀疏分解的图像修复方法

Image Inpainting Method Based on Sparse Decomposition

提出将冗余离散小波变换(RDWT)和波原子变换(WAT)作为字典对应用于图像稀疏形态成分分解,获得图像的结构与纹理;然后针对结构和纹理所具有的不同形态学特征,对结构采用具有曲率驱动、边缘强化和平滑去噪性能的CDD模型修复,对纹理采用基于样例的Criminisi纹理合成方法修复;最后合成获得修复结果。实验结果表明,该修复方法能够获得强且光顺的边缘,纹理清晰完整,相比于传统方法具有更好的修复结果。

This paper proposed a new couple of dictionaries which are redundant discrete wavelet transformation and wavelet atomic transformation,and applied it to image sparse morphological component decomposition to get the structure and texture.Then,based on the fact that the structure and texture have different characteristics,it uses curvature driven diffusion model which has curvature driven,edge enhancement and smooth denoising characteristics,and uses Criminisi texture synthesis method to inpaint the structure and texture respectively.At last,they are compounded and the inpainting result is got.The experiment results show that the new method can not only decompose the image very well,but also inpaint the image with strong and fairing edge,complete and clear texture.This method shows better results in image restoration compared to the classical ones.