Adaptive super-resolution reconstruction based on regularization
Super-Resolution (SR) reconstruction is referred to utilize information from a set of Low-Resolution (LR) images to reconstruct a High-Resolution (HR) image. Six parameters model for affine transformation is used in registration of LR image sequence, and SR reconstruction is performed according to the movement information of sub-pixel precision gained by the registration algorithm. In the reconstruction algorithm, fuzzy entropy is introduced as the judgment for the change of grayscale in image space based on the framework of regularization. Neighborhood homogeneous measurement of fuzzy entropy as the weighted coefficient of regularization item is adopted in the recognition of image edge and smooth region, and different regularization weights are used in different image regions, which are significant for maintaining sharpening in edge region and smoothing in flat area. Experimental results show that the proposed algorithm of SR reconstruction is effective, and the Peak Signal Noise Ratio (PSNR) is superior to the conventional regularization reconstruction algorithms.
image registration super-resolution reconstruction regularization fuzzy entropy
Qi-Shen Li Jun Guan Dan-Dan Zhang
School of Information Engineering Nanchang Hangkong University Nanchang, P.R. China
国际会议
太原
英文
370-373
2011-02-26（万方平台首次上网日期，不代表论文的发表时间）