Abstract: With the increasing application demands, image mosaic technology has become an important research topic in virtual reality technology, computer vision, computer graphics, video processing and other fields. This paper proposes a new image registration system based on the FAST feature. The system first uses FAST features to match both images, then uses the improved RANSAC algorithm to increase the registration accuracy. At last, it mosaics the both image by weighted fusion. Experiments show this system is more adaptive and stable. And it also not sensitive to rotation and affine transformation for the image. It can splice images which have overlap region. This system has a better speed performance than any other systems.

Key words: image mosaic, FAST feature, improved RANSAC, weighted fusion

摘要： 随着应用需求的增多，图像拼接技术已经成为虚拟现实技术、计算机视觉技术、计算机图形学以及视频处理等领域的一个重要研究课题。主要以FAST特征为核心，提出了一种基于FAST特征的快速图像配准系统。该系统首先通过FAST特征进行配准，然后通过改进的RANSAC算法增加配准准确率，最后通过加权融合完成图像拼接。实验表明，该系统有较好的适应性和稳定性，对图像的旋转、仿射变换均不敏感，能够较好地完成有重叠区域图像的拼接工作；同时该系统有较大的速度优势。

关键词: 图像拼接, FAST特征, 改进的RANSAC算法, 加权融合