自适应IQA阈值序列图像NLM超分辨重建方法

Adaptive IQA Threshold for NLM Multi-Frame Super-Resolution Method

序列图像超分辨率(super resolution,SR)算法可以利用多帧低分辨率图像之间的互补信息重建出一张高分辨率结果。传统非局部均值(non-local means,NLM)超分辨率重建方法的迭代次数选取和最佳SR重建结果筛选过程高度依赖使用者经验值和主观评价,这极大地增加了算法复杂度,降低了算法的鲁棒性。为了解决这两个问题,提出一种基于图像质量评价(image quality assessment,IQA)自适应阈值的NLM超分辨重建算法。通过设计一种SR重建结果质量评价指标,将该指标引入到NLM重建算法中:一方面作为阈值,用以确定算法迭代收敛条件;另一方面作为评价标准,用以筛选多个输出结果中重建效果最佳的高分辨率图像。实验结果表明,提出的算法能在有效保证鲁棒性的同时,极大地提升NLM超分辨率重建算法的运算效率。

Multi-frame image super-resolution(SR)algorithm uses complementary information between different low resolution images to reconstruct a high resolution image.Traditional SR method based on non-local means(NLM)is highly dependent on experience and subjective evaluation during the selection process of the iteration numbers and the best SR results.To solve these two problems,a NLM super-resolution method based on image quality assessment(IQA)adaptive threshold is proposed.The paper designs an evaluation index to assess the quality of SR result,and introduces this index to the NLM SR algorithm:on one hand,it functions as a threshold to determine the iterative convergence criteria of the algorithm;on the other hand,it helps selecting the SR result with the best reconstruction performance as assessment index.Several experiments prove that the proposed algorithm can ensure its robustness and greatly improve the computational efficiency of the NLM super-resolution reconstruction method.