一种结合像素空间信息和多维直方图的彩色图像快速分割算法

A Fast Color Image Segmentation Algorithm by Combining Pixel Space Information and 3D Histogram

本文提出了一种新的结合空间和彩色信息的彩色图像快速分割方法.首先对三个单色的直方图进行多阈值分割得到一组谷值和峰值,再用谷值向量将真彩色三维直方图分割成不同区域,然后以峰值组成种子向量实现三维直方图的多阈值分割,最后基于空间相邻关系的最优准则对分割的区域不连续性和"洞"重新分割,提高了过渡区域像素的分割精度.

By combining the pixel space information and color information,a fast color image segmentation algorithm is proposed in this paper to segment blood cell color images.The main ideas of this method are as follows：firstly,we apply the multi-threshold segmentation method to the three monochrome histograms to obtain a set of combinations of valley and peak values.Then,we utilize valley values to construct some hyperplanes in order to segment the true-color 3D histogram into different areas and employ the mean-shift algorithm with the seed vectors that are composed of peak values to implement the multi-threshold segmentation of 3D histograms.In order to solve the discontinuity problem and ＂hole＂ phenomenon of the areas obtained by single-color segmentation,a re-segmentation method using the optimal criterion based on adjacent space information is suggested.Some experiments are conducted and the results show that it can improve the segmentation accuracy to some extent.