一种鲁棒的图像局部特征区域的描述方法

A Robust Method for Local Image Feature Region Description

提出了一种鲁棒的图像局部特征区域的描述方法,即IWCS-LTP(Improved weighted center symmetric local trinarypattern)描述子.该方法对图像局部特征区域采用类似SIFT描述子的分块处理,可以使描述子包含更多的结构信息;采用ICS-LTP算子进行编码,可以在不大量增加描述子维数和计算量的同时对图像的梯度方向信息进行更具体的描述;采用加权纹理谱直方图计算方法可以使描述子包含图像的梯度幅值信息.大量的实验结果验证了该描述子的有效性.

In this paper,a robust method for local image feature region description,which is called IWCS-LTP（improved weighted center symmetric local trinary pattern）descriptor,is proposed.It uses a SIFT-like grid that makes the descriptor contain more structural information.By using the ICS-LTP operator,the descriptor can have more information of the image gradient direction without increasing the dimension of the descriptor and computing burden.This method uses the weighted texture spectrum histogram to construct the descriptor to contain the image gradient-magnitude information. The effectiveness of the designed descriptor has been validated by extensive experiments.