基于局部结构因子耦合双重度量规则的图像修复算法

Image inpainting algorithm based on local structural factor coupled with double metric rule

为了解决当前较多图像修复算法主要依靠置信度以及数据项来确定修复顺序,而忽略了待修复块的结构连续性,易导致修复的图像中存在不连续以及模糊等不良效应。对此,提出了一种基于局部结构因子耦合双重度量规则的图像修复算法。首先,通过像素点的方向导数构造了局部结构因子,利用局部结构因子以及置信度项建立了优先权度量函数,选取优先修复块。然后,通过像素点的二元函数构造了视觉优化模型,用于对样本块的尺寸进行自适应的调节。最后,利用图像块中像素值构造距离度量模型,通过图像块的灰度值构造相似度度量模型,将距离度量模型与相似度度量模型进行联合,建立双重度量规则,以距离度量值以及相似度度量为双重条件,搜索最佳匹配块。仿真实验结果显示,与当前图像修复算法相比,修复算法具有更高的修复质量与效率。

In order to solve the defects as the discontinuity and fuzzy effects induced by using the confidence and data items to determine the restoration sequence and ignoring the structural continuity of the repaired blocks in current image inpainting algorithms,A new image inpainting algorithm based on local structural factor coupling dual metric rules was proposed in this paper. Firstly,the local structure factor was constructed by directional derivative of pixels,and a priority measure function was established by using the local structure factor and confidence term to select the priority repair block. Then,the visual optimization model was constructed based on the two element function of pixel points to adaptively adjust the size of the sample block. Finally,the distance metric model was constructed by using the pixel value in the image block,and the similarity measurement model was constructed through the gray value of the image block which used to establish the dual metric rule to search for the best matching block under the conditions of distance metric and similarity measure. The simulation results show that this image inpainting algorithm has better inpainting quality and efficiency compared with the current image inpainting algorithm.