基于多分辨率技术的快速全景图图像匹配算法

为了对全景图像进行快速拼接和准确匹配，提出了一种新的基于多分辨率技术，利用相邻图像之间的颜色和边缘信息进行图像拼接的算法。目前应用于全景图制作的图像匹配算法通常要求拍摄时，相机要绕一垂直转轴（如三角架）做水平转动，同时不仅要求拍摄出的相邻图像有较多的重叠部分，而且相邻图像的光照差别不能太大，而新算法则避免了上述的对图像获取条件和图像质量的严格限制。因为该算法受相邻图像间光照强度变换和由镜头带来的图像形变的影响较小，所以在图像匹配实验中，该算法对各种条件下获取的图片都表现出很好的鲁棒性。实验结果表明，该算法在降低计算复杂度的同时，还有效地克服了由图像形变和不同光照条件所带来的图像的拼接困难。

A Multi-resolution Based Image Registration Method for Panorama

This paper presents a novel multi-resolution based approach which utilizes both the color and the edge information to create mosaic from two neighbor images. Unlike current image stitching algorithm, which usually requires pure horizontal camera panning, more overlapping and less exposure differences between two images, our method avoids the strict constraints on how the images are taken and the quality of the images. The proposed algorithm is robust because it is less influenced by the illumination differences and the distortion of the images brought in by lens. Experiments indicate that our algorithm is fast for the introduction of multi-resolution method.