資源描述:
《用于圖像重建的圖像快速高精度匹配算法研究分析》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在教育資源-天天文庫。
1、ResearchOnTheFastAndHighPrecisionMatchingAlgorithmForImageSuper-resolutionReconstructionAbstract用于圖像重建的圖像快速高精度匹配算法研究中文摘要圖像重建是計(jì)算機(jī)視覺研究領(lǐng)域的重要技術(shù),它是從多幅具有相互位移且有互補(bǔ)信息的圖像中獲得一幅更高分辨率圖像的過程。它通過對(duì)同一目標(biāo)拍攝多幅圖像,并利用圖像配準(zhǔn)算法求出由于攝像機(jī)抖動(dòng)引起的每幅圖像的位移量,最后利用重建算法提高圖像的分辨率。由于該項(xiàng)技術(shù)能大幅改善圖像質(zhì)量,因此在視頻監(jiān)控、醫(yī)學(xué)
2、診斷、遙感、天文觀測(cè)、生物信息識(shí)別等領(lǐng)域有很好的應(yīng)用前景。為了提升實(shí)際應(yīng)用中圖像重建速度,本文對(duì)快速配準(zhǔn)算法進(jìn)行了重點(diǎn)研究,研究?jī)?nèi)容主要包括特征提取、特征點(diǎn)定位、特征匹配算法及重建、基于OpenCV平臺(tái)實(shí)現(xiàn)等問題。本文主要工作如下:1.針對(duì)傳統(tǒng)圖像特征提取算法速度慢的問題,在分析對(duì)比各特征提取方法和策略的前提下,提出了一種快速的特征提取算法,該方法速度快、精度高,可以滿足準(zhǔn)實(shí)時(shí)圖像重建的要求。2.深入研究了Harris角點(diǎn)定位的原理,將角點(diǎn)的定位精度提升到亞像素級(jí),同時(shí)利用金字塔Lucas-Kanade算法進(jìn)行匹配以解決傳
3、統(tǒng)算法處理大而不連貫的運(yùn)動(dòng)時(shí)匹配不準(zhǔn)的問題,最后采用RANSAC算法去除偽角點(diǎn)對(duì),提高匹配準(zhǔn)確度。3.綜合分析比較各重建算法性能后,采用維納濾波去噪方法進(jìn)行圖像復(fù)原。4.基于OpenCV軟件平臺(tái),完成了圖像超分辨率重建算法程序,包括特征點(diǎn)提取、特征匹配、維納濾波。最后總結(jié)了本文所做工作,并對(duì)今后圖像配準(zhǔn)研究方向進(jìn)行展望。關(guān)鍵詞:圖像重建;亞像素;維納濾波;OpenCV;特征匹配ResearchOnTheFastAndHighPrecisionMatchingAlgorithmForImageSuper-resolution
4、ReconstructionAbstractResearchOnTheFastAndHighPrecisionMatchingAlgorithmForImageSuper-resolutionReconstructionAbstractImagereconstructionisanimportanttechnologyinthefieldofcomputervisionresearch,whichisaprocessthatputsmultipleimageswithdisplacementandcomplementary
5、informationintoahigherresolutionsingleimage.Firstly,thetechnologyobtainsmultipleimagesforthesametarget;secondly,itusestheimageregistrationalgorithmforcalculatingthedisplacementofeachimagecausedbycamerashake;finally,ahigherresolutionsingleimagecanbeobtainedbytherec
6、onstructionalgorithm.Becausethetechnologycansignificantlyimprovetheimagequality,Imagereconstructionhasagoodapplicationprospectinmanyfields,suchasvideomonitoring,medicaldiagnosis,remotesensing,astronomicalobservation,biologicalinformationidentificationandsoon.Inord
7、ertomeetthespeedofimagereconstructioninpracticalapplication,theresearchworkinthispaperfocusesonfastregistrationalgorithm,includingfeatureextraction,featurepointpositioning,featurematchingalgorithmandimplementationbasedonOpenCVplatform.Inthispaper,themaincontributi
8、oncanbesummarizedasfollows:1.Accordingtothelowspeedoffeatureextractionusingtraditionalalgorithms.Withananalysisofthefeatureextractionalgorithmsandstrate