基于k-mean聚类与灰度梯度最大熵的树木图像分割

Tree Image Segmentation Based on K-mean Clustering and Gray-gradient Maximum Entropy

提出一种基于k-mean聚类与灰度-梯度最大熵的树木图像分割算法,将要处理的树木彩色图像在RGB颜色空间下进行基k-mean聚类,通过选取合适的类参数实现初分割。由于灰度-梯度空间清晰地描绘图像中各个像素点的灰度、梯度的分布规律及图像目标与背景之间的边缘情况,采用灰度-梯度最大熵算法进行精分割,结合形态学后处理提取图像边缘最终将获得更理想的独立目标图像。与二维最大熵分割方法比较的实验结果表明,灰度-梯度最大熵算法提高了树木图像分割的准确度。

This paper proposed a k-mean clustering and gray-gradient maximum entropy image segmentation algorithm for trees. The color tree image was processed for base-k-mean clustering in the RGB color space, to achieve early segmentation parameters by se- lecting the appropriate class. Since gray-gradient space can clearly delineates between the distribution of edge cases each image pixel gray, gradient and image of the target and the background, the gray-gradient maximum entropy algorithm can be used to perform fine segmentation in combination with morphological processing of extracted image edge and the separated target image can eventually be ob- tained. Compared with the two-dimensional maximum entropy segmentation method, the results showed that gray-gradient maximmn entropy algorithm can improve the accuracy of image segmentation of trees.