基于凸集投影和线过程模型的超分辨率图像重建

Super-Resolution Image Reconstruction Based on Projections onto Convex Sets and Line Process Modeling

为了改善超分辨率重建图像的效果,提出了一种基于线过程模型的凸集投影方法.根据先验信息在图像重建中有重要作用的原理,该方法在数据一致性投影的基础上,将描述边缘信息的线过程模型作为图像的一个先验知识,用于重建过程中的平滑性约束投影,从而达到既保存图像的高频分量,又在一定程度上降低平滑区域噪声的效果.实验结果表明,该方法重建结果的视觉效果好于原方法,且信噪比有所提高,尤其是在高抽取率的情况下可以明显减少振铃效应.

In order to improve the super-resolution reconstruction of image, a method of projections onto convex sets (POCS) based on the line process modeling is proposed. According to the important role of the prior information playing in the image reconstruction, the line process modeling describing the edge information is considered as the prior information and used in the smoothness constraint projection. The experiment results demonstrate that the proposed method gives better supper-resolution results in the visual effect and higher signal-to-noise ratio than the original method. Especially.