面向图像语义分析的JSEG改进分割算法  

Segmentation for Image Semantic Analysis Based on Improved JSEG Algorithm

作　　者：周云[1] 张滢[1] 纪平[1] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《电视技术》2014年第1期183-186,共4页Video Engineering

摘　　要：JSEG算法是一种经典的被广泛应用于图像处理的图像分割算法,但是它存在着严重的过分割问题,因而不适用于对图像的语义分析处理。针对这个问题提出了两个方面的改进,一是改进了原算法中初始种子点的选取方法,二是在区域聚合过程中增加考虑了区域纹理特征的信息。实验表明,改进后的方法有效地改善了原算法中的过分割现象,分割后的图像区域更加符合对图像进行语义分析的要求。JSEG algorithm is a classic image segmentation algorithm which is used in image processing widely, but the over-segmentation problem is so severely that this algorithm isn＇t suitable for Image Semantic Analysis. Two improvements to solve this issue are proposed. One is improved the initial seed point selection method, the other is consider the information of texture feature in the region merging process. The experiments on large amounts of images show that the proposed method reduces the over-segmentation phenomenon effective and conforms to the requirements of image semantic analysis.

关 键 词：图像语义分析 JSEG算法 区域生长 区域聚合 

分 类 号：TN919.8[电子电信—通信与信息系统]