基于先验约束的图像超分辨率复原

SUPER-RESOLUTION RECONSTRUCTION OF IMAGE BASED ON PRIOR IMAGE CONSTRAINT

图像超分辨率复原是一个典型的不适定问题.近年来,借助先验信息改善图像超分辨率复原效果的方法已成为研究的一个热点,文章提出了一种利用先验图像的灰度分布作为约束的图像超分辨率复原的新方法.首先,利用最小鉴别信息构造复原图像与先验图像的约束,约束结果使得复原图像与给定的先验图像具有最相似的灰度分布;其次,基于置信策略给出自适应的正则化参数选取方法;最后给出实验结果,结果表明文中方法具有较好的复原效果和稳定性.

Super-resolution reconstruction of image is an ill-posed problem. Recently, super-resolution reconstruction based on prior information becomes a focus of research. In this thesis, a method using the gray-value distribution of prior image as constraint in super-resolution reconstruction of image was proposed. Firstly, the minimum information discrimination was used to construct the constraint of reconstructed image and prior image, which leaded to the reconstructed image similar to the prior characteristic image in gray-value distribution. Secondly, the method to choose regularization parameter adaptively based on fiducial strategy was given. The experimental results demonstrate that the method can provide better reconstructing and steady results.