【摘要】 使用Criminisi算法进行大区域纹理图像修复时会出现纹理延续错误的现象。针对该问题,提出一种改进的算法,主要体现在两方面:在修复顺序上,通过分析优先权计算方法,以置信度为主导改进优先权模型;在匹配策略上,结合几何距离关系,采用多块同时修复的策略,通过相似度函数选出各块的3个候选匹配块,引入几何距离修正因子确定各块的最佳匹配块。实验结果表明,该算法较好克服了原算法存在的优先权计算不准确以及匹配策略约束力不够导致的纹理延续修复错误的问题,与典型算法相比,其具有较好的视觉效果。更多还原

【Abstract】 The texture is propagated mistakenly when Criminisi algorithm is used to inpaint large region texture images.An improved algorithm was proposed to solve this problem,and the improvement was mainly embodied in two aspects.In the repairing order aspect,the priority model was optimized by analyzing the computational method of the priority using confidence value as the leading factor.In the matching strategy aspect,multi-block repairing strategy was adopted by considering the geometry distance.Three candidate blocks were chosen according to the similarity function combined with the distance correction factor synthetically.The best well matched blocks of each block were confirmed finally.The experimental results indicate that the proposed method overcomes the deficiency of the constraint of the matching strategy and the drawback of the error texture propagation.It has better visual effects compared with the state of the art algorithms.更多还原