一种基于BRISK改进的SURF红外图像拼接方法

摘要

在红外图像拼接技术的实际应用中,除了追求较高的拼接精度外,还应尽可能地提高拼接速度以满足实时性的要求.本文结合红外图像拼接技术的实际应用需求,将BRISK二值描述子和Canny边缘检测应用于SURF红外图像拼接,提出一种改进的SURF红外图像拼接方法.改进的红外图像拼接方法在特征点提取阶段将SURF特征点检测、Canny边缘检测和BRISK二值描述子有效结合,生成边缘SURF-BRISK特征点;在特征点匹配阶段采用自适应阈值的层次聚类方法作为搜索策略,并采用RANSAC算法剔除误匹配的特征点;在图像融合阶段采用渐入渐出的融合方法消除图像拼接缝隙.实验证明,本文所用红外图像拼接方法在对复杂场景下的红外图像进行拼接时,有效地提高了配准精度和拼接速度.%In the practical application of infrared image mosaic technology, in addition to the requirement of high stitching precision, the stitching speed should be increased as much as possible. In this paper, an improved SURF method, in which the BRISK descriptor and the Canny edge detection are applied to SURF feature extraction, is proposed based on the practical application requirements of infrared image mosaic technology. In the phase of feature extraction, SURF feature point detection is combined with Canny edge detection and the BRISK descriptor effectively to create the Edge SURF-BRISK feature. In the phase of feature matching, the threshold-adaptive hierarchical clustering method is adopted as the search strategy, and the RANSAC algorithm is used to eliminate the feature points of mismatching. In the phase of image fusion, the image gap is eliminated by a weighted fusion method. Experiments show that the infrared image mosaic method in this study can effectively improve the matching accuracy and stitching speed when stitching infrared images in complex scenes.