摘要： 利用传统凸集投影(POCS)算法进行图像超分辨率重建时，重建图像的边缘会产生振荡效应。针对该问题，提出一种改进的POCS算法。分析边缘振荡效应产生的原因，根据图像边缘两侧像素差值较大的特点，加权约束中心在边缘像素的点扩散函数，以减少边缘另一侧像素对当前像素的影响。实验结果表明，该算法能消除边缘振荡效应，提高重建图像的峰值信噪比。

关键词: 凸集投影算法, 超分辨率重建, 边缘振荡, 点扩散函数, 图像配准, 观测模型

Abstract: Traditional Project onto Convex Set(POCS) algorithm for image super-resolution reconstruction has the defect of edge oscillation. Aiming at this problem, an improved POCS algorithm is proposed. By analyzing the reason that causes the edge artifact, for the greater pixel difference of the both sides of edge, it weighted restricts the Point Spread Function(PSF) whose center is the edge pixel, and reduces the influence of the other side of pixels to the current edge pixel. Experimental results show that this algorithm can reduce the edge oscillation effectively and improve the Peak Signal to Noise Ratio(PSNR) of the reconstructed image.

Key words: Project onto Convex Set(POCS) algorithm, super-resolution reconstruction, edge oscillation, Point Spread Function(PSF), image registration, observation model

中图分类号: 

• TP391