基于GrabCut和HSV彩色空间的三七叶片病斑图像分割

Image Segmentation of Panax Notoginseng Leaf Diseases Based on GrabCut and HSV Color Model

针对三七叶部病斑图像中存在背景与病斑颜色相近和阳光照射下存在影子的情况导致对图像病斑分割精度不高、效果差的问题,提出一种基于GrabCut和HSV彩色空间模型的分割方法。通过GrabCut算法实现对复杂背景以及阳光照射下的影子与待分割三七叶片的分离。保留第一次分割结果的基础上,在HSV彩色空间下提出一种颜色距离的定义。根据定义的颜色距离,在使用颜色信息的同时结合空间信息,得到病斑的分割边界。实验表明,当色调分量因子k1和饱和度分量因子k2分别取0.1和0.05时得到的分割边缘效果较好。

Aiming at the problem that the background and the color of the lesion are similar to those of the lesion in panax notoginseng leaf image and the shadow in sunlight leads to the low accuracy and poor effect of the image lesion segmentation,a segmentation method based on GrabCut and HSV color space model is proposed.GrabCut algorithm is used to separate the shadows under complex background and sunlight from the leaves of panax notoginseng to be segmented.Based on the first segmentation result,a definition of color distance is proposed in HSV color space.According to the defined color distance,the segmentation boundary of the lesion can be obtained by combining the spatial information with the color information.Experiments show that when the hue component factor k1 and saturation component factor k2 are 0.1 and 0.05 respectively,the segmentation edge is better.