基于双边滤波的POCS超分辨率图像序列重建算法  

POCS super resolution reconstruction method for image sequences based on bilateral filter

作　　者：唐智飞[1] 禹晶[2] 肖创柏[1] 

机构地区：[1]北京工业大学计算机学院,北京100124 [2]清华大学电子工程系,北京100084

出　　处：《中国体视学与图像分析》2011年第1期84-88,共5页Chinese Journal of Stereology and Image Analysis

基　　金：北京市自然科学基金项目(4092006)

摘　　要：针对传统凸集投影(POCS)超分辨率图像序列重建算法使用高斯滤波器来估计点扩散函数(PSF)导致边缘模糊的现象,本文采用双边滤波器来估计PSF,双边滤波方法是将高斯函数和图像亮度信息相乘,优化后的权系数再与图像作卷积,使重建后的图像边缘得到很好保持。运动估计是POCS超分辨率图像序列重建算法的关键技术,本文选择结合图像金字塔的光流估计对图像序列进行配准,得到更加精确的估计结果。实验表明,可以使重建图像取得良好的视觉效果。The traditional projections onto convex sets（POCS） super-resolution image reconstruction method leads to the blur edge by using Gaussian filter,so we present bilateral filter instead of Gaussian filter to estimate the point spread function（PSF）.The bilateral filter can be got by Gaussian function multiplied with image brightness information.Finally the optimized weight coefficient makes convolution with the image so that the reconstructed image can keep the good edge.Motion estimation of image sequences is a key technology in the POCS super-resolution image reconstruction method.Optical flow estimation can get more accurate estimation results combining with image pyramid for image registration.The results of experiments show that the reconstructed image can achieve good visual effects.

关 键 词：超分辨率重建 POCS(凸集投影) 双边滤波器 光流估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]