段澤偉++秦雷
摘要:為更有效地對(duì)運(yùn)動(dòng)模糊圖像進(jìn)行圖像還原,本文在具有普遍意義的勻速直線運(yùn)動(dòng)退化模型下對(duì)點(diǎn)擴(kuò)散函數(shù)進(jìn)行分析,得到參數(shù)為模糊方向和模糊尺度構(gòu)成的二元函數(shù)。文章采用基于雙線性插值的方向微分算子和微分自相關(guān)算法對(duì)點(diǎn)擴(kuò)散函數(shù)進(jìn)行估計(jì)。實(shí)驗(yàn)表明,該方法可以對(duì)點(diǎn)擴(kuò)散函數(shù)的參數(shù)進(jìn)行高精度鑒別,采用經(jīng)過(guò)本文方法獲取的點(diǎn)擴(kuò)散函數(shù)均可實(shí)現(xiàn)還原算法對(duì)模糊圖像的有效還原。
關(guān)鍵詞:圖像還原,點(diǎn)擴(kuò)散函數(shù),方向微分算子,微分自相關(guān)算法
中圖分類(lèi)號(hào):TP18 文獻(xiàn)標(biāo)識(shí)碼:A 文章編號(hào):1009-3044(2016)06-0185-04
A Modified Analysis on the Motion Blurred Image
DUAN Ze-wei1 ,QIN Lei2
(1.AVIC Xi'an Aeronautics Computing Technique Research Institute, Xi'an 710056, China;2. Xi'an Communications Institute, Xi'an 710106, China)
Abstract: In order to effectively restore the motion blurred images, this paper analyzes the point spread function (PSF) under the uniform linear motion module and simplifies the PSF into binary function constituted of blur extent and blur direction. The paper adopts the direction of differential operator based on bilinear interpolation and differential autocorrelation algorithm to estimate the PSF. The experiments demonstrate that the method can precisely discriminate PSFs parameters , the restoration algorithms based on the estimated PSF obtained by the proposed method can effectively restore the experimental blur images.
Key words: Image Restoration; PSF; Direction of Differential Operator; Differential Autocorrelation Algorithm
圖像運(yùn)動(dòng)模糊現(xiàn)象是由于成像設(shè)備與觀測(cè)目標(biāo)之間的相對(duì)運(yùn)動(dòng)引起的在拍攝積分時(shí)間段內(nèi)場(chǎng)景能量在像平面上的非正常積累。[1]在現(xiàn)實(shí)使用中,需對(duì)模糊圖像復(fù)原,從而得到符合可視效果要求的圖像。運(yùn)動(dòng)模糊圖像在日常生活中普遍存在,近些年來(lái)關(guān)于運(yùn)動(dòng)模糊圖像的參數(shù)估計(jì)和復(fù)原處理成為研究焦點(diǎn)之一。圖像恢復(fù)可分為兩種情況:非盲圖像恢復(fù)和盲圖像恢復(fù)。在盲圖像恢復(fù)中,根據(jù)已有退化圖像估計(jì)出點(diǎn)擴(kuò)散函數(shù)以進(jìn)行恢復(fù)是重要的一類(lèi)方法。本文采用方向微分算子和微分自相關(guān)方法進(jìn)行模糊參數(shù)的估計(jì),以求得點(diǎn)擴(kuò)散函數(shù)(PSF)。為驗(yàn)證該方法所得PSF對(duì)還原的實(shí)用性和準(zhǔn)確性,本文調(diào)用三種常用的還原算法: 約束最小二乘方濾波法,盲卷積濾波法,Lucy-Richardson濾波法[2-4]。