基于样块改进的图像修复算法

Image Restoration Algorithm Based on Sample Block

图像修复是指对图像破损区域进行填充或者将图像中多余物体进行移除。在Criminisi算法的基础上进行改进,在待修复块优先权的计算过程中,由于等照度线的曲率可以反映图像的局部特征,块与块之间的方差值可以反映图像的边缘信息,因此,将二者考虑进来,确保修复过程能够准确有序地进行。在寻找最佳匹配块时,将等照度线的曲率也作为一个因素增加进来,有效地提高了最佳匹配块搜索的精确性。经过仿真实验证明,改进后的算法不仅在PSNR值上比原算法有所提高,而且修复结果也比原算法更加准确可靠。

Image inpainting refers to filling the damaged areas of the image or removing the redundant objects from the image. Conduct some improvements on the basis of Criminisi algorithm. As the curvature of equal illumination can reflect the internal characteristics of the image,variance value between the block and block can reflect the image edge information,adding them to the calculation of the priority of the repair block to ensure the inpainting can conduct accurately and orderly. When searching for the best matching block,the curvature of the equal illumination is added as a factor. The accuracy of searching for the best matching block is effectively improved. After the simulation experiment,the improved algorithm not only prove the PSNR value than the original algorithm,but also the obtained results are more accurate and reliable than the original algorithm.