自然生长状态下树叶图像的分割与提取

以用智能手机拍摄的自然生长状态下的含有复杂背景的树叶图像为研究对象，对图像的背景及其RGB3个颜色分量的特征进行分析，根据分析的结果提出采用超绿(EXG)算法和底帽变换算法相结合的方法对目标树叶进行分割. 对于绿色分量与其他2个分量差异大的背景采用EXG算法去除，而对于绿色分量与其他2个分量差异小的背景采用形态学的底帽变换来去除. 为了减小目标树叶分割的错分率，采用Otsu算法、形态学和边缘最大矩形对上述分割后的细节进行细化分割. 分割结果表明:文中所采用的算法可以很好地将目标树叶从背景中分割出来，错分率小于3.68%.

Leaves’ Image Segmentation and Extraction under Natural Growing Condition

The leaves obtained by mobile phones in complex backgrounds are taken as the research objects. According to the results of RGB components feature analysis, the target leaves can be segmented by the Extra-green character (EXG) and the bottom-hat transformation. For removing the background of the green component and the others with large difference, the bottom-hat transformation can be used. While for the other two components with small difference, the Extra-green character can be used. Moreover, in order to reduce the error rate of the segmentation, the Otsu algorithm can be used to modify its details. The result shows that all above of the algorithms can segment the leaves well from the complex backgrounds and all error rates are less than 3.68%.