对亮度和平移鲁棒的彩色全景图拼接算法

目的为了实现存在纵向错位和亮度差异的彩色全景图的拼接,提出一种基于HSV颜色空间的全景图拼接技术。方法提出了一种可变特征向量匹配算法,并且结合重叠区域对应特征匹配向量相关系数最大准则遍历搜寻对应的特征匹配向量,从而找到相邻子图的重叠区域,通过加权模板对重叠区域进行加权融合实现拼接。结果新算法对全景图的拼接效果优于常规算法,拼接质量Do EM值为0.999 d B,拼接平均耗时为499 ms,融合区域有渐进渐出的效果,有效解决了拼接错位和亮度突变问题。结论该算法对彩色全景图的拼接具有很好的适应性和可重复性,对子图的大小和颜色无严格要求。不仅拼接速度快,而且拼接精度高,是一套完整地针对存在纵向错位和亮度差异的彩色全景图的全自动拼接方案。

Stitching Algorithm for Color Panorama with Brightness and Translation Robustness

In order to realize the presence of longitudinal displacement and brightness difference, a panoramic image stitching technology was put forward based on HSV color space. This technology utilized a variable feature vector matching algorithm and conducted traversal of corresponding feature matching vectors by combining the maximum criterion of correlation coefficient for feature matching vectors of overlapping areas to find out the overlapping areas of adjacent images and achieve stitching by weight fusion for the overlapped area. This new algorithm had better stitching effect than traditional algorithm, with DoEM at 0.999 dB and average time consumption of 499 ms. The gradual fade-out effect effectively solved the problems of displacement and sudden change in brightness. This algorithm for stitching color panorama has good flexibility and repeatability and has no strict requirements on the size and color of the sub-graph. With fast stitching and high precision, this is a fully automatic stitching scheme for color panorama with longitudinal displacement and brightness difference.