Open AccessPolarization Image Registration Algorithm Based on Improved SIFT Descriptor
Author(s) -
Xiaodan Wang,
Xu Yang,
Chuanyi Qin
Publication year - 2019
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/677/3/032016
Subject(s) - sobel operator , scale invariant feature transform , artificial intelligence , computer science , computer vision , algorithm , edge detection , feature (linguistics) , feature extraction , image processing , pattern recognition (psychology) , image (mathematics) , linguistics , philosophy
With the application of polarization imaging technology more and more widely the research on how to obtain polarization information quickly and efficiently has always been the focus of people’s attention. In view of the numerous edge points generated by DoG (Difference of Gaussian) in the process of feature point extraction by SIFT(Scale-invariant Feature Transform) algorithm, which leads to tedious and time-consuming operation, an improved Sobel algorithm combined with SIFT operator image registration algorithm is proposed in this paper. The concrete implementation steps are as follows: firstly, the traditional two-direction Sobel algorithm is supplemented to eight directions, and then edge points are extracted. Then SIFT algorithm is used to detect feature points. Then, the feature points belonging to edge points are eliminated by Sobel operator. This method reduces the number of matching cycles. The processing results of RGB image and polarization image with this algorithm are compared with those of traditional SIFT algorithm, and the time consumption in the registration process is reduced by 60.5% on the premise of ensuring the accuracy.