目标尺寸先验与图像抽象的显著性检测
发布时间：2019-07-05 07:33
【摘要】：为了均匀突出显著目标并抑制小尺寸颜色独特的非显著目标对显著性检测的影响,研究了显著目标尺寸分布规律和多尺寸超像素抽象显著目标一致性,进而提出了基于目标尺寸先验的多尺寸超像素抽象显著性检测方法.该方法针对不同超像素尺寸抽象图计算图像的颜色独特性和颜色分布性,从局部和全局搜寻场景中的显著目标,并以尺寸分布规律指导显著目标的分割提取.通过对大量公开数据集中的图像进行测试的结果表明,该方法在有效检测显著目标的同时还能抑制小尺寸高对比度的非显著目标对显著性检测的影响,并生成均匀的高亮显著图.
文内图片：
图片说明：图１显著目标及其尺寸分布图抽象的尺寸选择和
[Abstract]:In order to uniformly highlight the significant target and suppress the influence of the non-significant target with unique small size color on the significance detection, the size distribution law of the significant target and the consistency of the multi-size super-pixel abstract significant target are studied, and then a multi-size super-pixel abstract significance detection method based on the target size priori is proposed. In this method, the color uniqueness and color distribution of images are calculated according to different super-pixel size abstract maps, and the significant targets in the scene are searched from local and global images, and the segmentation and extraction of significant targets is guided by the law of size distribution. The results of testing a large number of images in public data sets show that the method can not only effectively detect significant targets, but also suppress the influence of small size and high contrast non-significant targets on significance detection, and generate uniform highlighted images.
【作者单位】： 西安建筑科技大学信息与控制工程学院;西北工业大学电子信息学院;
【基金】：国家自然科学基金资助项目(61420106007)
【分类号】：TP391.41
本文编号：2510361
文内图片：
图片说明：图１显著目标及其尺寸分布图抽象的尺寸选择和
[Abstract]:In order to uniformly highlight the significant target and suppress the influence of the non-significant target with unique small size color on the significance detection, the size distribution law of the significant target and the consistency of the multi-size super-pixel abstract significant target are studied, and then a multi-size super-pixel abstract significance detection method based on the target size priori is proposed. In this method, the color uniqueness and color distribution of images are calculated according to different super-pixel size abstract maps, and the significant targets in the scene are searched from local and global images, and the segmentation and extraction of significant targets is guided by the law of size distribution. The results of testing a large number of images in public data sets show that the method can not only effectively detect significant targets, but also suppress the influence of small size and high contrast non-significant targets on significance detection, and generate uniform highlighted images.
【作者单位】： 西安建筑科技大学信息与控制工程学院;西北工业大学电子信息学院;
【基金】：国家自然科学基金资助项目(61420106007)
【分类号】：TP391.41
【相似文献】
相关期刊论文 前5条
1 黄小红;马云;陈曾平;;利用RCS序列估计空间目标尺寸的方法研究[J];信号处理;2005年06期
2 林庆;陈远祥;王士同;詹永照;;目标窗口尺寸自适应变化的Mean-Shift跟踪算法[J];计算机应用;2009年12期
3 罗山;黄欢;刘继红;丁大_8;;基于可变带宽Mean Shift的目标跟踪[J];微型电脑应用;2009年05期
4 刘旭东;郭东;白正勤;;利用尺寸特性的水面目标性质识别[J];计算机与数字工程;2011年12期
5 ;[J];;年期
，本文编号：2510361
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2510361.html
