一种新的基于特征区域的图像配准算法

全景拼图是近年来兴起的基于图像的绘制技术(IBR)中的一个重要研究方向。在全景视图的实现过程中,关键技术是实现重叠图像正确、平滑的无缝拼接。本文在原有图像配准技术的基础上,提出了一种新的基于特征区域的配准算法,该算法利用两幅重叠图像对应特征区域的相似性,经过特征选择、特征提取和特征匹配三个阶段,实现两幅图像的拼接。实验结果证明,该算法在拼接速度和效果方面的表现均令人满意。

A New Image Registration Algorithm Based on The Feature Area

Panorama mosaic is an important research area of Image-Based Rendering(IBR), which is popular in recent years. In the process of implementing panoramic image, on of the key techniques is to match the overlapping images correctly each other, making that the stitched image be seamless and smooth. Based on the analyses of the existing algorithms, this paper introduces a new feature-area based registration algorithm by computing the similarity of the feature area between the two images. The mosaic of two images can be implemented through three steps: feature select, feature distill and feature match. The results of experiment show the efficiency of the algorithm.