一种灰度级抽取的图像分割方法研究

Gray Level Extraction Method for Image Segmentation

阈值分割方法依据图像直方图分析,利用最优化法则进行最优阈值选取,完成图像分割,这类方法将直方图最优分析等价于图像最优分割.但是,直方图提供的是图像灰度级统计信息而无空间关系信息,则阈值分割可能会导致分割失败,所以有必要进行灰度级空间信息的考察以完成图像分割任务.提出一种灰度级抽取分割方法,首先利用层次聚类方法对各灰度级像素进行空间近邻考察,并给出一种类别数与类间距关系确定最优类别数的方法,然后对近邻灰度级进行抽取并组合,最终形成分割结果.本文对该方法与若干阈值分割方法进行了对比验证,实验表明,所提方法在图像分割方面具有优势,能够得到较为满意分割结果.

Traditional threshold segmentation method relies on the analysis of the histogram of an image, an optimal threshold is ob- tained through a specified optimization algorithm with respect to the histogram, then the segmentation result is achieved. This method- ology regards the optimal histogram analysis as the best image segmentation. However, a histogram merely supplies the statistical in- formation of the gray level rather than the spatial relationship among them, thus, the threshold based segmentation probably leads to segmentation failure, and then, the investigation on the spatial relationship among gray levels is necessary in order to complete the segmentation task. This paper presents a spatial estimation of gray levels using hierarchical clustering techniques, and builds a relation between cluster numbers and cluster distances to determine optimal cluster number, then extracts and combines the neighborhood gray levels, and finally gives the segmentation result. Comparison between the proposed method and threshold based method is conducted, the experiment shows that the given method prevails and more comfortable results can be obtained.