基于改进SURF算法的图像拼接技术研究

Technologies of Image Mosaic Based on Improved SURF Algorithm

针对传统的加速稳健特征(SURF)算法在图像拼接过程中计算复杂度高以及匹配精度不佳等问题,提出一种基于SURF的改进算法,首先基于加速分割检测特征(FAST)算法快速提取图像特征点,利用SURF算法对提取到的特征点进行特征描述,然后通过改进的k-d树最近邻查找算法(BBF)寻找图像间的匹配点,与双向匹配的自适应阈值配准法相结合进行图像的匹配,利用改进的随机抽样一致性(RANSAC)算法对提取的特征点进行误匹配剔除,最后使用渐入渐出的加权融合算法对图像进行拼接。实验表明与传统的SURF+RANSAC算法相比,本文算法的图像拼接速度快,匹配精度更高。

Aiming at the problem of high computational complexity and poor registration accuracy image mosaic,a new image mosaic method based on improved speed up robust feature(SURF)was proposed.Firstly,feature points were extracted by the accelerated segment test algorithm and described based on the SURF descriptor.Secondly,the matching points between images were searched by using the improved k-d tree nearest neighbor search algorithm(best bin first).Then the adaptive threshold registration algorithm of bidirectional matching strategy was used for image matching.Finally,the random sample consensus(RANSAC)algorithm was used to eliminate false matching points,After that,the image mosaic was conducted based on the incremental weighting fusion algorithm.Experimental results show that the proposed method is more efficient and accurate than traditional image mosaic methods based on SURF and RANSAC.