基于多种色彩空间的像素覆盖分割方法

Pixel Coverage Segmentation Method Based on Multiple Color Spaces

基于线性分离的像素覆盖分割方法需要人为确定分类数量,覆盖分割矩阵的初始化也未能很好的利用图像中已有的信息.为了解决这些问题,提出一种基于多种色彩空间的像素覆盖分割方法.首先使用HSI空间峰值直方图确定图像分割数,其次利用FCM聚类方法确定端元矩阵,在不同色彩空间定义不同的距离度量对覆盖分割矩阵进行初始化,再利用像素覆盖分割方法对图像进行分割.最后采用PRI和VOI评价指标对在不同色彩空间的像素覆盖分割结果进行评价,根据评价结果确定图像的最优分割色彩空间和最优分割.实验结果表明,该方法分割结果优于FCM聚类方法,通过PRI和VOI评价指标可以为图像分割选择合适的色彩空间.

The coverage segmentation based on linear unmixing needs to determine the clustering number by hand, and the coverage segmentation matrix initialization failed to make good use of the image information. To solve these problems,a pixel coverage segmentation method based on multiple color spaces is proposed. Firstly,the peaks of the image histogram in the HSI space is used to determine the segmentation number. The end-members are determined by the FCM clustering method. Different distance metrics are defined in different color space and they are used to initialize the coverage segmentation matrix. Then the original pixel coverage segmentation is applied to image segmentation. The PRI and VOI values are used to evaluate the effectiveness of the coverage segmentation results in different color spaces, and determine the best segmentation color. Experiment results show that the proposed segmentation method is better than the FCM clustering method,and the best segmentation color space could be determined by the PRI and VOI values.