A Robust Method for Constructing Rotational Invariant Descriptors
Yi Yang,Fajie Duan,Ling Ma,Jiajia Jiang
DOI: https://doi.org/10.1016/j.image.2017.10.010
IF: 3.453
2017-01-01
Signal Processing Image Communication
Abstract:Unlike most existing descriptors that only encode the spatial information of one neighborhood for each sampling point, this paper proposed two novel local descriptors which encodes more than one local feature for each sampling point. These two local descriptors are named as MIOP (Multi-neighborhood Intensity Order Pattern) and MIROP (Multi-neighborhood Intensity Relative Order Pattern), respectively. Thanks to the rotation invariant coordinate system, the proposed descriptors can achieve the rotation invariance without reference orientation estimation. In order to evaluate the performance of the proposed descriptors and other tested local descriptors (e.g., SIFT, LIOP, DAISY, HRI-CSLTP, MROGH), image matching experiments were carried out on three datasets which are Oxford dataset, additional image pairs with complex illumination changes, and image sequences with different noises, respectively. To further investigate the discriminative ability of the proposed descriptors, a simple object recognition experiment was conducted on three public datasets. The experimental results show that the proposed local descriptors exhibit better performance and robustness than other evaluated descriptors.