自适应阈值分割与局部背景线索结合的显著性检测
发布时间：2018-08-14 11:49
【摘要】：为了提高显著性算法对不同类图像的适用性以及结果的完整性,该文提出一种基于自适应阈值合并的分割过程与新的背景选择方法相结合的显著性检测算法。在分割过程中,生成相邻区块的RGB以及LAB共六通道融合的颜色差值序列,采用区块面积参数的反比例模型生成自适应阈值与颜色差值序列进行对比合并。在背景选择过程中,根据局部区域背景-主体-背景的相对位置关系线索,得到背景区域,再对结果进行边缘优化。该算法与其它算法相比得到的显著图不需要外接其他阈值算法即生成二值图,自适应阈值合并能排除复杂环境中的物体细节,专注于同等级大小物体的显著性对比。
[Abstract]:In order to improve the applicability of salience algorithm to different kinds of images and the integrity of the results, this paper proposes a salience detection algorithm based on adaptive threshold combining segmentation process and new background selection method. In the process of segmentation, the RGB and LAB color difference sequences of adjacent blocks are generated, and the adaptive threshold and color difference sequences are generated by using the inverse scale model of block area parameters. In the process of background selection, the background region is obtained according to the clues of the relative position relationship between the local region background and the subject background, and then the edge of the result is optimized. Compared with other algorithms, the salient map obtained by this algorithm does not need to be added to other threshold algorithms to generate binary graphs. Adaptive threshold merging can eliminate the details of objects in complex environments and focus on the salience comparison of objects of the same size.
【作者单位】： 河北工业大学电子信息工程学院;河北工业大学计算机科学与软件学院;
【基金】：天津市科技计划项目(14RCGFGX00846,15ZCZDNC 00130) 河北省自然科学基金面上项目(F2015202239)~~
【分类号】：TP391.41
本文编号：2182767
[Abstract]:In order to improve the applicability of salience algorithm to different kinds of images and the integrity of the results, this paper proposes a salience detection algorithm based on adaptive threshold combining segmentation process and new background selection method. In the process of segmentation, the RGB and LAB color difference sequences of adjacent blocks are generated, and the adaptive threshold and color difference sequences are generated by using the inverse scale model of block area parameters. In the process of background selection, the background region is obtained according to the clues of the relative position relationship between the local region background and the subject background, and then the edge of the result is optimized. Compared with other algorithms, the salient map obtained by this algorithm does not need to be added to other threshold algorithms to generate binary graphs. Adaptive threshold merging can eliminate the details of objects in complex environments and focus on the salience comparison of objects of the same size.
【作者单位】： 河北工业大学电子信息工程学院;河北工业大学计算机科学与软件学院;
【基金】：天津市科技计划项目(14RCGFGX00846,15ZCZDNC 00130) 河北省自然科学基金面上项目(F2015202239)~~
【分类号】：TP391.41
【相似文献】
相关期刊论文 前2条
1 张伟;隋青美;;基于多小波的局部背景隐马尔可夫模型图像去噪[J];青岛科技大学学报(自然科学版);2008年06期
2 ;[J];;年期
，本文编号：2182767
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2182767.html
