摘要： 对凸集投影(POCS)图像超分辨率重建算法中的残差修复阈值选取问题进行分析，提出一种基于图像超分辨率重建的改进算法。改变传统POCS算法中固定残差修复阈值模式，通过引入低分辨率图像清晰度这一先验信息，用表征图像相对清晰度的参数控制阈值，从而实现整个重建过程阈值动态选取。实验结果证明，该算法能提高残差计算和阈值选取的针对性，对提升重建图像的清晰度及信噪比具有较好的效果。
关键词:
超分辨率重建,
凸集投影,
残差修复阈值,
加权凸集投影,
图像平均梯度,
点扩散函数,
空间频率
Abstract: This paper aims at the selection of residual-repaired threshold in image super-resolution reconstruction based on Projection Onto Convex Sets(POCS) and improves it. Changing the mode of fixed residual-repaired threshold in the traditional POCS algorithm, this paper introduces a prior information about low resolution image definition into POCS to control the threshold by characterization parameter of image relative definition and realizes the dynamic threshold selection in the entire progress of reconstruction. Experimental results show that the improved algorithm has very good effect on improving the clarity of the reconstructed image and PSNR.
Key words:
super-resolution reconstruction,
Projection Onto Convex Sets(POCS),
residual-repaired threshold,
weighted POC,
image average gradient,
point spread function,
spatial frequency
中图分类号:
姚琦, 沈松, 朱飞. 基于加权POCS的图像超分辨率重建[J]. 计算机工程, 2013, 39(3): 264-266,271.
TAO Qi, CHEN Song, SHU Fei. Image Super-resolution Reconstruction Based on Weighted POCS[J]. Computer Engineering, 2013, 39(3): 264-266,271.