结构优化的彩色图像稀疏表示的修复方法

Color Inpainting by Structural Optimization and Sparse Representation

针对彩色图像修复过程中,RGB颜色模型三通道间的相关性以及其结构复杂性导致修复效果不理想的缺点,提出了将修复过程从RGB颜色模型转化到YUV颜色模型进行,采用稀疏表示的方法来提高彩色图像的修复效果.首先将图像从RGB颜色模型映射到YUV颜色模型并对不同成份进行分块,然后利用FastICA算法分类训练获得相对应的完备字典,最后利用优先权函数约束待修复块的修复顺序,并结合重构算法SL0对修复块进行重构,实现了图像的修复.实验表明,运用该方法能较好地修复条状破损、小块破损以及大块破损的图像和文字移除,改善了传统修复方法在边缘修复时出现的边缘断裂或者纹理延伸现象.在修复时,利用优先权函数来决定修复优先度,使边界和纹理的修复更加符合人眼视觉效果.

Since the correlation and the complexity structure of RGB color model leads to the shortcomings on non-ideal effect of the color image restoration process, this paper proposes a new sparse representation color inpainting method to improve the repair effect. Firstly, the RGB color image is mapped into YUV model and divided parts. Then useful information is extracted from the volume of natural and undamaged images. Using the FastICA algorithm to obtain a complete dictionary; at last priority function is used to determine to repair the turn to be repaired, combining with reconstruction algorithm SL0 to treat repair piece of reconfiguration and repaint. Experimental results show that proposed method can effectively repair the strip breakage, small damage, large pieces of broken images and the text removed, and repairing boundary and the texture is more conformed to the human eye vision effect.