机器视觉中的尺度可控图像分割方法

摘要

机器视觉中的尺度可控图像分割方法是一种基于几何流的超像素生成法.该方法是在待处理图像上放置粒子,然后粒子根据曲线演化理论进行扩张,粒子的边界曲线则将图像划分为多个面积大致相同的超像素子区域.该方法一方面能够通过度量粒子内外的差异性来严格控制粒子边界曲线的扩张,从而避免出现过分割与欠分割现象,另一方面随着超像素子区域数量的增加,对图像中目标区域的细节处理效果更加显现,对图像处理效率的显著性也逐渐突出.%The scale-controlled image segmentation in machine vision is a super-pixel generation method based on geometric flow. This method places the seeds on an image to be treated, and these seeds are then expanded according to the theory of curve evolution, while boundary curve of the seeds divides the image into multiple super-pixel sub-regions with roughly the same area. On the one hand, the method can strictly control the expansion of the boundary curve of seeds by measuring the difference between seeds inside and outside, thus to avoid the phenomenon of over segmentation or under segmentation, and on the other, with the increase of super-pixel sub-regions, treatment effect on the details of target area in the image becomes more obvious, processing efficiency on the image gradually prominent.