摘要
针对传统特征点浮点数描述符存储空间需求高、匹配速度慢的不足,提出了一种新的二进制串特征点描述符构造方法.该方法首先将特征点的局部图像区域划分为多个子区域,再通过比较子区域内部的像素灰度和子区域间的平均灰度来构造二进制串描述符,该描述符具有存储空间小、匹配速度快的显著优点.最后在标准图像库上进行了实验测试.实验结果表明,该描述符取得了相当或优于BRIEF描述符、SIFT描述符的性能,且具有明显的速度优势,适用于对速度和存储空间要求较高的图像匹配任务中.
Aimed at the defects of high storage space requirements and insufficient matching speed of the tra- ditional feature points＇ floating-point descriptors, this paper presents a novel binary descriptor. In the proposed method, the paper firstly partitions the local image region extracted by feature detectors into several sub-regions, and then constructs a binary sequence descriptor by comparing the gray values in each sub-region and the average gray degree between sub-regions. The proposed binary descriptor has significant advantages of both memory-sav- ing and time-saving in point matching applications. Finally the paper performs experimental tests on popular stan- dard image library and the result demonstrates that the proposed descriptor achieves a matching performance simi- lar or superior to BRIEF and SIFT, and even runs much faster, which makes it suitable for the image matching ap- plication with relatively higher requirements on speed and storage space.
作者
曲智国
谭贤四
邓斌
万凡兵
QU Zhiguo, TAN Xiansi, DENG Bin, WAN Fanbing(Air Force Early Warning Academy, Wuhan 430019, Chin)
出处
《空军预警学院学报》
2018年第2期127-131,141,共6页
Journal of Air Force Early Warning Academy
关键词
图像特征点描述符
二进制串
特征点匹配
image feature points＇ descriptor
binary sequences
point matching
