仪器仪表学报
儀器儀錶學報
의기의표학보
CHINESE JOURNAL OF SCIENTIFIC INSTRUMENT
2014年
8期
1900-1906
,共7页
虹膜块状纹理%组合窗口搜索%K-means聚类
虹膜塊狀紋理%組閤窗口搜索%K-means聚類
홍막괴상문리%조합창구수색%K-means취류
iris plaque-like texture%combined window searching%K-means cluster
目前的虹膜纹理特征提取方法无法获得特定类型纹理的大小、位置和形态信息,而这些信息同样是纹理特征的重要组成部分。将虹膜纹理分为块状、线状和环状3类,以其中的块状纹理为检测对象,提出一种利用组合窗口搜索实现虹膜块状纹理检测的方法。该方法针对块状纹理的灰度分布特征定义一组尺寸可变的组合窗口,利用这些窗口搜索预处理后的虹膜图像,找到满足块状纹理匹配条件的区域,再利用聚类方法对这些区域进行二值化,实现虹膜中块状纹理的检测。该算法对图库中人工标定的918处坑洞和127处色素斑的检测正确率分别为92.16%和76.38%,实验结果表明,该方法能够从背景复杂、多种形式纹理共存的可见光虹膜图像中检测出块状纹理,这些纹理信息将有助于提高虹膜识别的鲁棒性。
目前的虹膜紋理特徵提取方法無法穫得特定類型紋理的大小、位置和形態信息,而這些信息同樣是紋理特徵的重要組成部分。將虹膜紋理分為塊狀、線狀和環狀3類,以其中的塊狀紋理為檢測對象,提齣一種利用組閤窗口搜索實現虹膜塊狀紋理檢測的方法。該方法針對塊狀紋理的灰度分佈特徵定義一組呎吋可變的組閤窗口,利用這些窗口搜索預處理後的虹膜圖像,找到滿足塊狀紋理匹配條件的區域,再利用聚類方法對這些區域進行二值化,實現虹膜中塊狀紋理的檢測。該算法對圖庫中人工標定的918處坑洞和127處色素斑的檢測正確率分彆為92.16%和76.38%,實驗結果錶明,該方法能夠從揹景複雜、多種形式紋理共存的可見光虹膜圖像中檢測齣塊狀紋理,這些紋理信息將有助于提高虹膜識彆的魯棒性。
목전적홍막문리특정제취방법무법획득특정류형문리적대소、위치화형태신식,이저사신식동양시문리특정적중요조성부분。장홍막문리분위괴상、선상화배상3류,이기중적괴상문리위검측대상,제출일충이용조합창구수색실현홍막괴상문리검측적방법。해방법침대괴상문리적회도분포특정정의일조척촌가변적조합창구,이용저사창구수색예처리후적홍막도상,조도만족괴상문리필배조건적구역,재이용취류방법대저사구역진행이치화,실현홍막중괴상문리적검측。해산법대도고중인공표정적918처갱동화127처색소반적검측정학솔분별위92.16%화76.38%,실험결과표명,해방법능구종배경복잡、다충형식문리공존적가견광홍막도상중검측출괴상문리,저사문리신식장유조우제고홍막식별적로봉성。
At present,the iris texture feature extracting methods can not acquire the size,location and morphology characteristics of spe-cific type of textures;however,this information is also a significant part of iris texture features.In this paper,the iris textures are classi-fied into three kinds:plaque-like,linear and circular textures.An iris texture detection method based on combined window searching is presented,which aims at the plaque-like iris textures.A series of variable size combined windows are defined based on the gray distribu-tion of the plaque-like textures.These windows are adopted to search the preprocessed iris images and find the regions in which the plaque-like textures exist.And then,with K-means cluster method,the regions are converted to binary form and the plaque-like textures are detected.To test this method,the method was used to detect 918 crypts and 127 plaques that were artificially marked in our gallery;and the detection accuracies were 92.16%and 76.38%,respectively.The experiment results show that the proposed method can detect the iris plaque-like textures from the visible iris images with complex background and multiple texture coexisting.This texture information will be helpful for improving the robustness of iris recognition.