【摘要】 提出一种基于尺度不变特征变换特征点的全自动球面全景图生成算法,将原始图像投影到同一个虚拟球面上,再对球面展开图融合拼接,以避免叠加误差;并设计了一种非匹配点剔除方法,比常用的RANSAC算法提高了效率和准确性.实验验证了该算法的稳健性和快速性.更多还原

【Abstract】 An algorithm based on scale invariant feature transform(SIFT)is presented.After extracting key points,a new method is introduced to remove unmatched key points,which is showed to be more efficient than random sample consensus(RANSAC)method.Images are projected onto a virtual sphere and the remapped images are stitched by a multiresolution method.The experiments verify our algorithm’s efficiency and robustness.更多还原