摘要
利用灰度共生矩阵提取图像纹理特征值,然后根据熵值的大小来实现模块大小的选择。另外,在寻找最佳匹配块时,同时考虑了颜色信息的差异和空间距离的因素。最后,给出了客观评价图像修复质量的PSNR度量。实验表明,与Criminisi算法相比,该方法得到的修复效果更自然,更符合人的视觉感知。
This paper firstly utilizes gray-level co-occurrence matrix to extract image＇s texture feature, then, it completes the choice of module size according to the value of entropy. In addition, when finding the best match, it considers the differences of color information and the factor of spatial distance at the same time. Finally, it gives PSNR measurement to evaluate image inpainting quality objectively. The experiments show that, compared with Criminisi algorithm, the inpainting effects got from this method are more natural and more fitful to people＇s vision.
出处
《微型机与应用》
2012年第21期44-46,49,共4页
Microcomputer & Its Applications
基金
福建省教育厅基金资助项目(JB07023)
关键词
图像修复
纹理合成
灰度共生矩阵
熵
模块大小
最佳匹配块
image inpainting
texture synthesis
gray-level eo-oceurrence matrix
entropy
module size
best match
