动态调节耦合纹理差异量化的图像修复算法

Image inpainting algorithm based on dynamic adjustment coupling texture difference

为解决当前图像修复算法存在的块效应以及模糊效应等不足,提出动态调节机制耦合纹理差异的图像修复技术。利用权重配比,构造优先权度量模型,评估待修复区像素点的优先权;利用样本块的梯度特性,构造结构差异度量函数,计算样本块及其邻域块之间的差异;根据度量结果,设计动态调节机制,调整样本块大小;按照调整后样本块的大小,利用像素点的R、G、B颜色值搜索最优匹配块;利用像素点的梯度值,构造纹理差异度量模型,联合最优匹配块,完成图像修复。实验结果表明,与当前图像修复算法相比,所提算法具备更好的图像视觉质量与鲁棒性能。

To solve these defects such as block effect and fuzzy effect existing in current image inpainting algorithm,an image inpainting algorithm based on dynamic adjustment mechanism coupling texture difference was proposed.The priority measurement model was constructed based on weight matching method to measure the priority of the pixels in the repair area.The structure difference measure function was constructed using the gradient property of the sample block to measure the difference between the sample block and its neighborhood block.The dynamic adjustment mechanism was used to adjust the size of the sample block.According to the size of the sample block after adjustment,the optimal matching blocks were searched using the R,G and B color values of the pixels,so as to repair the sample blocks.The model of texture difference measurement was constructed using the gradient value of the pixels,which can be used to update the relative reliability and complete the image restoration.Experimental results show that,the proposed algorithm has better visual effect and robust performance compared with the current image inpainting algorithm.