文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition
侵权投诉

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

sieves, to extract eye gesture parameters from the gray-level data directly. The proposed method resembles the work Automatic recovery of eye gestures from image se- done by Matthews et al. [11] for lip reading, but differs in quences is one of the importa

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

Figure1.TrackedlefteyeoftheClairese-quence

3.SieveDecomposition

Diffusionbasedscale-spaceapproaches[10]inimageprocessingallows(i)tosimplifytheimagesatmultiplescales,byremoving nedetails,and(ii)toaccuratelyesti-matethescale,thepositionandtheidentityoftheobjectsinanimage,frommeasuresofextremaandedges.However,simpli cationisnotonlyachievedbydiffusion,butalterna-tivenonlinearmorphological ltershavebeenintroducedtothisend[2],[3].Here,weconsidertheonedimensionalin-creasingscaleofalternatingsequential lterswith atstruc-turingelements,knownassieves.Thesievesdecomposeaonedimensionalboundedfunction,,toasequenceofin-creasingscale:granulefunctions,

,representingthekeyfeaturesoftheoriginalsignal.

Aformalpresentationofthedecompositionisgivenasfollows:

(1)

Theoperatorcanbeanopening,aclosing,anopening/closing,oraclosing/opening,fora atstructuringelementofsize.Notethatfortheonedimen-sionalcasetheopeningandclosingoperatorsarede nedas:

with,thesetofintervalsofwithelements.Thesievesareabletorobustlyrejectnoiseinthemannerofmedians,anddosomoreeffectivelythandiffusionbasedschemes.Moreoverthesievesarealsoshownexperimentallytobefast[2].

AsitcanbeseenfromEq.1andthede nitionoftheop-erator

extremaofaparticularscaleareremovedateachstageofthecascade.Thedifferencesbetweensuccessive,calledgranulefunctions,arelocallynon-zeroforintervalsofsamples.Thenon-zerointervalsarecalledgranules.Thegranulesarescalerelatedprimitivestowhichtheedgeinformationisintimatelybounded[3].

4.Eyegestureparametersestimation

Theonedimensionalrecursivesievede nedabovemaybeusedtodecomposeatwodimensionalimageofaneye,byscanningverticallyovereachimagecolumn,i.e.inthedirectiononwhicheyemotionoccurswhenblinking.Theresultinggranularityspectrumcontainstheposition,ampli-tudeandscaleinformationofthesetofgranulesthatde-scribetheimage.Byignoringpositionandamplitudein-formationMatthewsetal.[11]de nedascalehistogramasthenumberofgranules,thegranulesenergy(squaredam-plitude),thegranulesamplitudesorthegranulesabsoluteamplitudesummedacrosstheentireimage.Thishistogramwasused,foraudiovisualspeechrecognition,asafeaturevectorforstandardHiddenMarkovModels.Insteadofworkingwiththeeigenvaluedecompositionapproach[11],wesuggestanotherwayofcombiningthescales.

FollowingthesamebasicideaasMatthewsetal.[11],weform,foreachscale,aspatio-temporalcurvewherethegranulesamplitude,summedacrosstheentireeyeimage,arerepresentedasafunctionoftime(frame).

Eye Activity Detection and Recognition Using Morphological Scale-Space Decomposition

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email