资源描述:
《基于突变信号检测的光学标记识别图像分割方法》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、基于突变信号检测的光学标记识别图像分割方法摘要:针对无定位信息的光学标记识别(omr)图像填涂区的精确定位问题,提出了一种基于小波变换突变信号检测的图像分割方法。该算法首先计算图像的水平和垂直投影函数,然后投影函数经过迭代小波变换后检测其突变点,突变点能够精确地反映omr信息的边界位置。检测算法的适应性基于有限次数的小波变换和突变信号检测过程。实验结果表明该算法具有较高的分割精度和稳定性,分割精度均方差可以达到0.4167个像素。而且由于算法只使用图像的水平和垂直投影信息,因此具有较高的执行效率;投影函数的统计特性和小波变换的多分辨特性则使
2、得该分割算法对噪声不敏感。关键词:光学标记识别;小波变换;图像分割;突变点检测;多分辨分析omrimagesegmentationbasedonmutationsignaldetectionmalei1,2,liujiang1,2*,lixiao.peng1,2,chenxia1,21.shandongengineeringresearchinstituteforimageacquisitionandprocessing,jinanshandong250101,china;2.shandongshandaoumasoftc
3、ompanylimited,jinanshandong250101,chinaabstract:aimingattheproblemofaccuratepositioningofopticalmarkrecognition(omr)imageswithoutanypositioninformation,animagesegmentationapproachbasedonwavelettransformationmutationssignaldetectionisproposed.firstlythehorizontalandverticalp
4、rojectiveoperationareprocessed,thenthesefunctionsaretransformedbywaveletanddetectedpointmutation,thesepointsareabetterdescriptionoftheboundaryofomrinformation.thisalgorithmadaptabilitybasedonlimitedtimesofwavelettransformandmutationssignaldetection.theexperimentalresultsdem
5、onstratethatthemethodpossesseshighaccuracyofsegmentationandstability,themeansquareerrorofsegmentationaccuracycanbe0.4167pixels.theprocessingofthismethodwasefficientbecausethesegmentationonlybasedthehorizontalandverticalinformation.thisalgorithmwasnotsensitivetonoisebecauseo
6、fprojectionfunctionsstatisticcharacteristicandmulti-resolutionanalysisofwavelet.concerningtheaccuratepositioningofopticalmarkrecognition(omr)imageswithoutanypositioninformation,animagesegmentationapproachofmutationsignaldetectionbasedonwavelettransformationwasproposed.first
7、ly,thehorizontalandverticalprojectiveoperationswereprocessed,andthenthesefunctionsweretransformedbywavelettodetectmutationpoints,whichcanbetterreflecttheboundaryofomrinformation.thisalgorithmsadaptabilityisbasedonlimitedtimesofwavelettransformandmutationsignaldetection.the
8、experimentalresultsdemonstratethatthemethodpossesseshighaccuracyofsegmentationands