Transformed self-exemplars based SR algorithm doesn't make the best use of the texture and edge information of images,and moreover,PatchMatch used in the algorithm was easy to get trapped in suboptimal local minima.Focusing on the problem,this paper proposed a new self-similarity based SR algorithm using optimized PatchMatch.Firstly,it used simulated annealing algorithm to improve PatchMatch's ability to escape from local minima.Secondly,it defined edge structure similarity factor and introduced it into the evaluation function of patches similarity.Finally,inspired by the statistical prior called local self-similarity,this paper searched similar patches in the position coordinate space satisfying Gaussian probability distribution.Experiments show that compared with other state-of-the-art SR algorithms,the proposed method makes progress in both visual effect and objective evaluation.%针对基于自示例几何不变性的图像超分辨率算法没有充分利用图像的纹理和边缘信息,且采用的PatchMatch图像块匹配算法容易陷入局部极小点的问题,提出了一种基于改进PatchMatch的自相似性图像超分辨率算法.首先,利用模拟退火算法提高PatchMatch跳出局部极小点的能力;然后,定义边缘相似度因子并将其引入到图像块相似性的评价函数中;最后,受局部自相似性统计先验的启发,采用服从高斯概率分布的位置坐标搜索空间进行相似图像块匹配.实验表明,与当前先进算法相比,所提算法在视觉效果和客观评价指标上都有一定的提高.
展开▼
