摘要
使用传统的特征描述方法SIFT在单一尺度上描述图像特征会丢失一部分重要信息,影响图像的正确匹配结果。为了解决这一问题,本文在多尺度模糊空间内提取特征描述子。信息熵从图像显著性角度估计特征点及其周围的信息,能获得更多的图像关键内容,本文提出了基于局部熵的图像特征描述方法。首先,在高斯差分空间(DOG)内计算特征点的多层SIFT描述子,同时统计特征点在每层尺度上的局部熵,计算特征点在每层的熵值占所有层熵总和的百分比,利用所得百分比与每层描述子做乘积;然后,累加所有层描述子;最后,使用平方根算法得到最终局部熵特征描述子。通过与其他描述子的对比实验结果可知,本文提出的局部图像描述方法在精确-召回率、平均均匀准确度和正确匹配数方面具有强鲁棒性。
The traditional model, local scale-invariant features, does not capture some parts of the important information of an image, which will lead to bad matching result. In order to address this problem, extracting feature information from different scales in blurring space is better than the classic algorithm. A local entropy method can estimate interesting point information from its surrounding, which can obtain more image content. This paper provides a new image feature description based on local entropy. First, the orient histograms from different scales within different Gaussians space are computed, and each local entropy value is estimated from each scale level. Then, all of the descriptors of the same interesting point are fussed based on the ratio of each entropy to the whole local entropy value of the same feature in different scale levels. Experiment results demonstrate that, compared with order out-of-state local descriptor, the two descriptors provided in this paper have strong robust under different conditions.
出处
《吉林大学学报(工学版)》
EI
CAS
CSCD
北大核心
2017年第2期601-608,共8页
Journal of Jilin University:Engineering and Technology Edition
基金
国家自然科学基金项目(61101155)
吉林省发展和改革委员会产业创新专项项目(2016C035)
关键词
计算机应用
图像特征检测
图像特征描述
局部熵
computer application
image feature detection
image feature description
local entropy