改进的最大相似度区域目标分割算法  

Improved object segmentation algorithm based on maximal similarity region

作　　者：徐杰[1] 许存禄[1] 邢磊[1] 马晓滢[1] 

机构地区：[1]兰州大学信息科学与工程学院,甘肃兰州730000

出　　处：《计算机工程与设计》2015年第9期2519-2523,共5页Computer Engineering and Design

基　　金：国家自然科学基金青年基金项目(60903101);兰州大学中央高校基本科研业务费专项基金项目(lzujbky-2011-62)

摘　　要：针对最大相似度区域融合交互式图像分割(maximal similarity region merging,MSRM)算法在目标与背景存在模糊区时分割效果不理想的缺陷,提出一种改进的MSRM算法。利用导向滤波算法对图像进行细节增强,在不丢失边缘信息的特点的情况下,对原模糊图像进行增强去除模糊区,结合颜色分割算法过滤增强图像的大量冗余信息,提高目标分割精度,扩大MSRM的应用范围。仿真结果表明,改进算法实现了对模糊图像目标更准确、更快速的分割。Aiming at the defect of unsatisfactory target extraction of interactive image segmentation by maximal similarity based region merging （MSRM）, an improved algorithm of MSRM was proposed. Guided image filtering character was used to enhance the original image and remove the fuzzy area without missing edge information, thus expanding the application range of MSRM. The color segmentation was combined to filter the redundancy information, thus improving the segmentation accuracy. The simulation results show that the proposed algorithm realizes accurate and fast object segmentation.

关 键 词：最大相似度 模糊区 导向滤波 图像增强 颜色分割 冗余信息 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]