结合简单交互和标记分水岭的复杂背景叶片图像分割方法  

SEGMENTATION METHOD FOR LEAF IMAGE UNDER COMPLEX BACKGROUND COMBINING SIMPLE MAN-MACHINE INTERACTION AND MARKER-BASED WATERSHED SEGMENTATION

作　　者：高理文[1] 林小桦[2] 罗晓牧[1] Gao Liwen;Lin Xiaohua;Luo Xiaomu(School of Medical Information Engineering, Guangzhou University of Chinese Medicine, Guangzhou 510006, Guangdong, China;School of Chinese Materia Medica, Guangzhou University of Chinese Medicine, Guangzhou 510006,Guangdong, China)

机构地区：[1]广州中医药大学医学信息工程学院,广东广州510006 [2]广州中医药大学中药学院,广东广州510006

出　　处：《计算机应用与软件》2016年第8期211-215,241,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61301294);广东省教育科学"十二五"规划教育信息技术研究专项(13JXN021);广州中医药大学青年英才项目(QNYC20140203)

摘　　要：复杂背景叶片图像分割是当前的研究热点和难点。提出一种结合简单交互和标记分水岭的方法，有效地解决了该问题。首先，让用户在复杂背景叶片图像上奇偶相间地标定叶边缘点和叶外点；经过一系列处理得到标记图像。接着，把复杂背景叶片图像转换为灰度图像、L-a-b-图像；以标记图像为参数，分别对灰度图像、a-分量图像、b-分量图像，进行标记分水岭分割。最后，以投票方式综合得到最终的分割结果。对20个种类,200张复杂背景叶片图像的分割实验表明：该方法能实现较准确的分割，并且能保留叶片细节部分;对个别叶片有细小的错分，但对其形状影响轻微。Segm enting le a f images w ith com plex background is cu rre n tly a d iffic u lt and foca l p o in t o f research. In this p a p e r, we propose am ethod w h ich com bines sim ple m an-m achine in te ra ctio n w ith m arker-based watershed segm entation to solve the problem o f le a f imagesegm entation e ffe ctive ly. F ir s t, the m ethod lets the user m ark the p e riphe ral and e xte rio r points o f a le a f in odd-even sequence on the le a fimage u n der com plex background ; after a series o f processing the m arked image is produced. S e co n d ly, the m ethod transform s the le a f imageu n der com plex background in to greyscale image and L * a * b * im a g e ,then ap plies the m arker-based watershed segm entation to greyscaleim a g e , a * com ponent image and b * com ponent image re spectively w ith the previous m arked image as param eters. F in a lly , the fin a lsegm entation re su lt is achieved com prehensively by means o f vo tin g . The segm entation experim ents on 200 le a f images in 20 categories undercom plex backgrounds in d ica te that this m ethod can realise precise seg m entation, and can preserve the de tailed parts o f a le a f. There are tin ym istakes in some segm entations, b u t have slig h t im p a ct on le a f shapes.

关 键 词：复杂背景 叶片 图像分割 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]