基于纹理特征与最优稀疏表示的图像修复算法

目的解决当前图像修复算法忽略了对修复块后续的优化处理,导致修复图像易出现不连贯效应以及块效应等的不足。方法提出基于纹理特征与稀疏表示的图像修复算法,首先利用像素点对应的数据项,构造了优先权模型。然后,利用像素点在R,G,B分量上对应的像素值来构造纹理特征度量模型,对待修复块中像素点对应的纹理特征进行度量,并根据度量结果,选择其对应样本集的大小。引入SSD型,从样本集中搜索与待修复块最相似的最优样本块,对待修复块进行填充。最后,利用最优样本块函数,构造最优稀疏表示模型,从而实现图像修复。结果仿真结果显示,与当前图像修复算法相比,所提图像修复算法具备更高的复原质量,能有效克服修复图像中出现的不连贯效应以及块效应。结论所提算法具有较高的修复视觉质量,在数字图像处理领域具有较好的应用价值。

Image Inpainting Algorithm Based on Texture Features and Optimal Sparse Representation

The work aims to solve the problem that the image inpainting algorithm ignores the subsequent optimization of the patch, leading to the lack of coherent effect and block effect in the inpainting image. An image inpainting algorithm based on texture features and sparse representation was proposed. Firstly, the priority model was constructed with the data item corresponding to the pixel points. Then, the texture feature measurement model was constructed with the pixel values corresponding to the pixel points on components R, G and B to measure the texture features corresponding to pixel points in the block to be restored. Moreover, the size of its corresponding sample set was selected according to the measurement results. Through the introduction of SSD model, the optimal sample block that was the most similar to the block to be repaired was searched from the sample set and used to fill the block to be repaired. Finally, the optimal sparse representation model was constructed with the optimal sample block function to achieve image inpainting. The simulation results showed that, compared with the current image inpainting algorithm, the proposed image inpainting algorithm had higher recovery quality and could effectively overcome the incoherent effect and block effect in inpainting. The proposed algorithm has higher visual quality of inpainting and better application value in the field of digital image processing..