基于导向滤波器的超像素分割

Superpixel Segmentation Based on Guide Filter

针对大多数超像素分割算法均通过迭代逐步逼近图象的边缘,虽多次迭代后分割线能够到达对象的边缘,但计算效率低、重合度较低,文中提出了一种强化对象边缘的方法。利用导向滤波器对图像进行预处理,使对象的边缘进一步强化;对图像进行超像素分割促使像素快速聚类。试验证明了提出算法的可行性。

For most of the superpixel segmentation algorithms,edge of image is gradually approximated by iteration.Although the segmentation line can reach the edge of the object after multiple iterations,the computational efficiency is low and the coincidence degree is low. In this paper,a method to strengthen the edge of the object is proposed.The image is preprocessed by the guide filter,and the edge of the object is further strengthened. The image is segmented to make the pixel fast clustering. The experiment proves the feasibility of the proposed algorithm.