一种基于图像区域分块的SIFT快速配准方法

A Fast SIFT Registration Method Based on Image Region Segmentation

针对传统尺度不变特征变换(SIFT)算法在特征提取与描述时计算量大、实时性差的问题,提出一种基于区域分块的SIFT的快速配准方法。首先,将匹配图像和待匹配图像分割成若干均匀的子图,通过计算每个子图的信息熵值与设定阈值比较来确定局部子图的特征类型;对筛选出来的特征区域的子图进行特征提取和生成PCA-SIFT描述子,对筛选出来的平坦区域直接跳过,不进行检测。实验结果表明:提出的方法在保证配准精度90%以上的情况下,计算时间减少了15%~25%左右,提高了图像配准的速度。

For a large amount of computation and poor real-time performance of the traditional scale invariant feature transform（SIFT） algorithm in feature extraction and feature description, a fast registration method based on region segmentation SIFT is proposed. First, the matching image and the image to be matched are divided into several uniform subgraphs. The feature type of the local subgraph is determined by calculating the information entropy value of each subgraph and setting the threshold value. The features of the subgraph are extracted and the PCA-SIFT descriptor is generated. Then, the fiat area screened out is skipped directly and not detected. Experimental results show that the proposed algorithm in ensuring the accuracy of registration of more than 90% of the cases, the calculation time is reduced by about 15%- 20%, and the speed of image registration is improved.