基于图像结构张量的改进样本图像修复算法

Improved Sample Image Inpainting Algorithm Based on Image Structure Tensor

由于Criminisi算法针对复杂结构待修复图像的修复效果不好,提出一种基于图像结构张量和改进Criminisi样本的图像修复算法,采用图像结构张量和图像梯度的组合对图像结构进行检测,以实现显著和非显著图像结构的区分。对Criminisi算法中的优先级函数进行改进,提出一种最佳距离启发式方法进行优先权重计算,充分考虑了待修复像素点周围的图像特征,从而增强了图像结构的影响。实验结果表明:所提算法的修复性能优于其他算法,能够有效实现图像修复。

The image restoration strategy based on sample Criminisi algorithm is not very effective when repairing images with complex structures.To solve this problem,this paper proposes an image restoration algorithm based on image structure tensor and improved Criminisi sample.The algorithm uses the combination of image structure tensor and image gradient to detect the image structure to achieve the distinction between significant and non-significant image structures.In addition,the priority function in the Crinisi algorithm is improved,and an optimal distance heuristic method is proposed to calculate the priority weight,which takes full account of the image features around the pixels to be repaired,thus enhancing the image structure.Experimental results show that the proposed algorithm can effectively achieve image restoration,and the restoration performance is better than other algorithms.