基于边缘盒与低秩背景的图像显著区域检测算法
发布时间：2019-08-16 14:31
【摘要】：针对现有显著性区域边界不明确和检测效果鲁棒性较差等问题,提出了一种新颖的图像显著区域检测方法,该方法结合了边缘盒粗定位和低秩背景模型细筛选来提高显著区域的检测性能。首先,对基于边缘盒的图像显著区域检测方法进行改进,采用OTSU方法自适应计算边缘模值的最佳分割阈值,以替代固定分割阈值,降低边界点检测误差;其次,在基于边缘盒检测到的可疑显著区域上,采用鲁棒主成分分析方法获取图像的低秩分量,构建背景模型,并基于背景差分方法剔除背景区域,减少显著区域的虚检现象。在PASCAL VOC 2007数据集上的实验结果表明,提出的方法明显提高了显著区域检测的精确度和召回率,同时具有较高的检测效率。
[Abstract]:In order to solve the problems of unclear boundary of salient region and poor robustness of detection effect, a novel image salient region detection method is proposed, which combines edge box rough location and low rank background model fine screening to improve the detection performance of significant region. Firstly, the image salient area detection method based on edge box is improved, and the OTSU method is used to calculate the optimal segmentation threshold of edge modulus adaptively, in order to replace the fixed segmentation threshold and reduce the detection error of boundary points. Secondly, in the suspicious significant area based on edge box detection, robust principal component analysis (PCA) is used to obtain the low rank component of the image, the background model is constructed, and the background region is eliminated based on the background difference method to reduce the false detection of the significant area. The experimental results on PASCAL VOC 2007 data set show that the proposed method obviously improves the accuracy and recall rate of significant area detection, and has high detection efficiency.
【作者单位】： 江苏师范大学计算机学院;太原理工大学电气与动力工程学院;中国矿业大学计算机科学与技术学院;
【基金】：江苏省教育科学“十二五”规划课题(C-c/2011/02/010) 江苏省教育科学“十二五”规划2013年度立项课题(D/2013/02/273)的阶段性成果 山西省重大专项项目(20131101029)资助
【分类号】：TP391.41
[Abstract]:In order to solve the problems of unclear boundary of salient region and poor robustness of detection effect, a novel image salient region detection method is proposed, which combines edge box rough location and low rank background model fine screening to improve the detection performance of significant region. Firstly, the image salient area detection method based on edge box is improved, and the OTSU method is used to calculate the optimal segmentation threshold of edge modulus adaptively, in order to replace the fixed segmentation threshold and reduce the detection error of boundary points. Secondly, in the suspicious significant area based on edge box detection, robust principal component analysis (PCA) is used to obtain the low rank component of the image, the background model is constructed, and the background region is eliminated based on the background difference method to reduce the false detection of the significant area. The experimental results on PASCAL VOC 2007 data set show that the proposed method obviously improves the accuracy and recall rate of significant area detection, and has high detection efficiency.
【作者单位】： 江苏师范大学计算机学院;太原理工大学电气与动力工程学院;中国矿业大学计算机科学与技术学院;
【基金】：江苏省教育科学“十二五”规划课题(C-c/2011/02/010) 江苏省教育科学“十二五”规划2013年度立项课题(D/2013/02/273)的阶段性成果 山西省重大专项项目(20131101029)资助
【分类号】：TP391.41
【相似文献】
相关期刊论文 前10条
1 梅跃松;杨树兴;莫波;;一种实时抗噪的运动区域检测算法[J];红外与激光工程;2006年S4期
2 包莹莹;沈梦伟;;基于运动区域检测的黑板设计[J];现代计算机(专业版);2009年08期
3 景慧昀;韩琦;牛夏牧;;显著区域检测算法综述[J];智能计算机与应用;2014年01期
4 万梦玉;张梦琪;;基于游戏中区域检测碰撞的改进措施[J];计算机光盘软件与应用;2014年08期
5 王艳娟;陈晓红;;图像感兴趣区域检测技术[J];科技资讯;2006年32期
6 王艳娟;陈晓红;黄晓欣;;图像感兴趣区域检测技术[J];计算机与数字工程;2007年05期
7 陈硕;于晓升;吴成东;陈东岳;;用于场景分类的显著建筑物区域检测[J];上海交通大学学报;2011年08期
8 肖创柏;司薇;邓米克;王首道;;非感兴趣区域时空分析的运动显著区域检测[J];北京工业大学学报;2012年01期
9 蒋鹏;秦小麟;刘s，
本文编号：2527495
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2527495.html
