多目标显著性区域提取算法  

Salient Region Extraction Algorithm for Multi-target

作　　者：孟琭[1] 陈妹雅 MENG Lu;CHEN Mei-ya(School of Information Science&Engineering,Northeastern University,Shenyang 110819,China)

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110819

出　　处：《东北大学学报（自然科学版）》2018年第10期1380-1384,共5页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金资助项目(61101057)

摘　　要：结合对象估计和超像素分割,提出面向多目标的显著性区域提取算法.首先,应用对象估计对图像中的多目标作初步检测,得到若干个显著性区域的初步结果;然后,再将这些显著性区域与超像素分割的结果作图像拼接,完善这些显著性区域;最后,将图像拼接的结果二值化,作为多目标显著性区域提取的最终结果.结果表明:所提算法可实现面向多目标的显著性区域提取.与3个经典算法的比较结果表明:所提算法在面向多目标显著性区域提取时更优.Combining object estimation and super-pixel segmentation,a salient region extraction algorithm for multi-target was proposed.First,object estimation was used to make a preliminary extraction of multi-target in image,and the preliminary results of several salient regions were obtained.Then,these several salient regions were concatenated with the results of super-pixel segmentation to complete the saliency extraction.Finally,the concatenated regions were binarized as the final results of salient region for multi-target.The results showed that the proposed algorithm can achieve multi-target salient region extraction.The comparison with three classical algorithms indicated that the proposed algorithm is better when it is faced with salient region extraction for multi-target.

关 键 词：多目标 显著性区域 对象估计 超像素分割 图像处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]