小波域图像块修复  

Image Repairing Research Based on Block-by-Block in Wavelet Domain

作　　者：陈永彬[1] 张松[1] 

机构地区：[1]成都电子机械高等专科学校通信工程系,成都610031

出　　处：《数据采集与处理》2009年第5期650-654,共5页Journal of Data Acquisition and Processing

摘　　要：描述了一种在小波域中分块图像修复的新算法。该算法将修复过程分为两个步骤;首先寻找小波细节系数,用小波系数的能量来确定每一个图像梯度向量的大小;然后使用图像梯度向量来决定修复区中的哪个块应该首先被填充,使用基于小波域的纹理合成方法来具体填充这个块。连续迭代执行这两个步骤,直到修复区完全被填满。实验结果表明该算法的有效性。A novel algorithm is developed for block-by-block image repairing in the wavelet domain. The image repairing is carried out in two critical steps. Firstly, wavelet detail coefficients are optimized, and the magnitude of the image gradient vector is determined by the energy of the wavelet coefficients. Then, the image gradient vector is utilized to decide whether the block in the impairing region is to be filled first and such block is then filled based on the adapted method for texture synthesis in the wavelet domain. Two steps are performed in a nested way until the repairing region is completely filled. Experimental results show that the algorithm is effective.

关 键 词：图像修复 小波 纹理合成 梯度向量 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]