结合图论的JSEG彩色图像分割算法

摘要

Joint Systems Engineering Group ( JSEG) is a classical method of image segmentation algorithm. It fully takes the local image information into account,so it can get more precise segmentation boundary. But the JSEG algorithm has the large computation and over-segmentation problems. For this reason,propose a segmentation algorithm combining JSEG and graph theory. Firstly,calculate J value on-ly on a small scale instead of the iterative process on multi-scale. Secondly,use the K-means clustering method on the J-map to get over-segmentation regions. Finally,use a point to replace a region,and then use the graph theory for region merging. Experimental results show that the new algorithm has the advantage of a high accuracy and low complexity.%静态图像压缩标准( JSEG)分割算法是一种经典的图像分割方法，它充分考虑到了图像的局部信息，可以获得比较精确的分割边界。但JSEG算法在分割过程中计算量相当大并且分割结果容易出现过分割现象。由此，文中提出一种结合图论的JSEG图像分割算法。首先去除JSEG算法中在多个尺度上反复计算J值的过程，改为仅在一个小尺度上进行计算。其次，在得到的J图上使用K-means方法进行聚类，分割得到过分割区域。最后，将分割后的小区域对应为图中的点，进而利用图理论的方法进行区域合并。实验结果表明新算法具有高精度和低复杂度的优势。