鱼眼图像协同性目标检测方法
Fish-eye lens image co-saliency detection algorithm
张力丹 1朱均超 1冯为嘉1
作者信息
- 1. 天津理工大学自动化学院天津市复杂工业系统控制理论及应用重点实验室,天津300384
- 折叠
摘要
由于鱼眼图像存在着严重非线性畸变,使得对图像中的信息进行准确识别变得尤为复杂,目前的显著性检测算法往往只针对单张图像进行检测,显著性测度少,检测结果误差较多,协同性目标检测算法利用多幅鱼眼图像之间的协同关联,对多幅鱼眼图像同时进行显著性信息检测,增加了显著性检测约束条件,通过构建全局关联性信息约束,对鱼眼图像中不同畸变程度的显著性物体进行有效的提取,弥补了单张图像检测中的不足,提高了检测结果的准确性.
Abstract
Because of serious nonlinear distortion in fish-eye lens image,it's hard to detect information clearly.Almost detection algorithm devote to single image significant detection.Lower detection measurement but more deviation.fish-eye lens image co-saliency detection algorithm has been put forward,take advantage of cooperativity between multi-image,detect multi-fish-eye lens images simultaneously,add saliency detection measurement,construct global corresponding constraints,all of those make it more accuracy to detect saliency information in uncorrected fish-eye lens images.
关键词
鱼眼图像/信息提取/显著性检测/协同性Key words
fish-eye lens Image/information extraction/saliency detection/cooperativity引用本文复制引用
基金项目
国家自然科学基金(61172185)
天津市高等学校科技发展基金(20100705)
出版年
2017