方差约束因子耦合搜索区域判定模型的图像修复算法

为了解决Criminisi算法在图像修复过程中无法保证修复块的优先级顺序,从而导致修复质量不佳的问题,提出了方差约束因子耦合搜索区域判定模型的图像修复算法.首先,将待修复块分割为两个子块,通过子块的方差构建方差约束因子,并利用方差约束因子改进Criminisi算法中的优先权函数;然后,在二维直角坐标系中对损坏区域进行测量,根据测量结果选取损坏基准值,以构建搜索区域判定模型,确定最优匹配块的搜索范围;最后,引入SSD(Sum of Squared Differences)模型在搜索区域中选取最优匹配块,利用最优匹配块中像素点与待修复块中对应像素点的像素差值构造置信度更新模型,对置信度进行更新,实现图像的修复.实验结果表明,与其他图像修复算法相比,本文算法具有更好的图像修复视觉质量.

Image Inpainting Algorithm Based on Variance Constrained Factor and Search Region Decision Model

In order to solve the defect as poor visual quality induced by not guaranteeing the priority order of repair blocks in the process of image inpainting in Criminisi algorithm. An image inpainting algorithm with variance constraint factor and search region decision model has been proposed in this paper. Firstly, the repaired blocks are divided into two sub blocks, and the variance constraint factors are constructed by the variance of the blocks. The priority function of the Criminisi algorithm is improved with variance constraint factor of blocks. Secondly, the damaged area is measured in the two-dimensional Cartesian coordinate system, and the damage reference value is selected according to the measurement results for constructing the search area decision model to determine the search range of the optimal matching block. And lastly, the SSD model is introduced to select the optimal matching block in the search area. The confidence update model is constructed by using the pixel difference between the pixels in the optimal matching block and the corresponding pixels in the patch to be restored, and then the image inpainting is realized. Experimental results show that, compared with the current image inpainting algorithm, the proposed algorithm has better visual effects.