基于形态学尺度空间的多尺度图像分割研究

Multiscale Image Segmentation Based on Morphological Scale Space

给出了图像层次分割的新架构,阐述了基于重构形态学操作的尺度空间产生方法,提出了基于形态学尺度空间和形态学分水岭的多尺度分割策略,其目的是生成图像的分割堆栈,并通过链接技术获得有意义的对象分割。该文给出了对各种图像的分割的实验结果,验证了分割策略的有效性。

This paper presents a new framework for the hierarchical segmentation of various images. The scale-space is based on morphological anisotropic diffusion that uses reconstruction morphological operators. The proposed scheme for multiscale segmentation comprises a morphological scale-space and morphological gradient watersheds. Its aim is to produce steak of image segmentation and obtain a meaningful hierarchy among the objects in the image by linking technique. The experimental results of the proposed scheme are obtained with a wide range of image including natural scenes. The result shows that the scheme is effective.