【摘要】 全景图拼接是实现虚拟现实场景漫游的一种重要手段,本文利用SUSAN检测算子抽取图像的角点特征,然后用快速RANSAC算法配准图像,计算出图像之间的单应性矩阵,同时通过KLT跟踪,提高投影矩阵的精度。最后在图像融合中,采用像素值加权的方法。实验结果表明,文章提出的方法能有效提高配准精确性,并具有鲁棒性。更多还原

【Abstract】 Panoramic image mosaic is an important method of travelling in virtual reality environment.This paper uses SUSAN corner detector to extract the corner’s characteristic from the images,then register the images by the PERANSAC algorithm,and compute the homography matrix between the two images.And,the KLT trace is also used to refine the homography matrix.At last,Pixel weighted method is used to blend the overlapping area of image.The results shows that the accuracy of registration is improved greatly and the method is robust.更多还原