Chapter 6 - Feature Extraction

Chapter 6 - Feature Extraction

ID:40056501

大小:4.07 MB

页数:45页

时间:2019-07-18

Chapter 6 - Feature Extraction_第1页
Chapter 6 - Feature Extraction_第2页
Chapter 6 - Feature Extraction_第3页
Chapter 6 - Feature Extraction_第4页
Chapter 6 - Feature Extraction_第5页
资源描述:

《Chapter 6 - Feature Extraction》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、§6-FeatureExtractionDr.DirkSchniedersDepartmentofComputerScienceTheUniversityofHongKongLevelsofProcesses★Typically,threetypesofprocessescanbeclassifiedinthiscontinuum:★low-levelprocesses,★mid-levelprocesses,and★high-levelprocessesImageComputerProcessi

2、ngVisionlow-levelmid-levelhigh-levelprocessesprocessesprocesses2013-2014COMP75022Mid-LevelProcesses★Mid-levelprocesses★Higherleveltaskse.g.imagesegmentation,imagedescription,classificationandrecognition★Inputsaregenerallyimages★Outputsareattributesext

3、ractedfromtheimagese.g.edges,contours,identityoftheindividualobjectsmid-levelimageprocessattributes2013-2014COMP75023ImageInterpretation★Theintensityvalueofeachpixeldependsonanumberoffactors(1)Thenatureanddistributionoflightsources(2)Theshapeofthestru

4、cturesinthescene(3)Thereflectancepropertiesofthesurfaces(4)Thepositionofthecamera(5)Thepropertiesofthelensandtheimagesensor★Typically,computervisionapplicationsaimatdeducingthestructuresinthescene(2)andthepositionandpropertiesofthecamera(4,5)fromtheim

5、ageindependentoftheothers2013-2014COMP75024WhyFeatureExtraction?★Eachimagecontainsahugeamountofdata★E.g.,1920*1080*8bits★Theaimoffeatureextractionistodramaticallyreducetheamountofdataby★preservingtheusefulinformationintheimages(e.g.,thoserelatedtocame

6、raposeandscenestructures)★discardingtheredundantinformationintheimages(e.g.,thoserelatedtoreflectanceandlightingconditions)★Allsubsequentinterpretationisthenperformedonthegenericrepresentation,nottheoriginalimage2013-2014COMP75025SalientFeatures★Edgea

7、ndcornerfeaturesarecommonlyusedincomputervision★Theyarefairlycheaptocompute★Theyrequirerelativelysmallstoragespace★Theyaresomewhatinvarianttochangesinlighting★Theyprovidestrongcluesthatcanhelpthevisionprocess2013-2014COMP75026ImageStructure1.2.3.2013-

8、2014COMP75027ImageStructure-Featureless1.Afeaturelessregionischaracterizedbyasmoothvariationofintensities2013-2014COMP75028ImageStructure-Edge2.Apatchcontaininganedgerevealsanintensitydiscontinuityinonedirection2013-2014COMP75029ImageStructure

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。