图像校正和插值是基于柱面全景图像视图合成中的关键问题。首先分析了极线采样方法在柱面图像校正像素保持率和像素有效率上的优势,通过计算极线的数量、范围及起止位置,有效保留了源图像的信息,减少了以往算法中由于使用单应性映射而导致的弱采样和过采样效应;其次,在新视点插值上,根据柱面全景图像的成像模型,推导了合成视图上像素点的位置和颜色计算公式。最后给出了基于仿真图像和实际图像的视图合成结果。
View synthesis framework based on cylindrical panoramic images is proposed,and image rectification and interpolation of cylindrical panoramic images are mainly investigated.Advantage of epiline sampling in rectification of cylindrical image is analyzed,and reference images are sampled efficiently and sufficiently by calculating the start-end position,number,range of epiline.The approach of epiline sampling on cylindrical images is adopted in order to reduce image deformation,object distortion and resolution losing due to perspective transformation in some other algorithms.For novel view interpolation,according to the imaging model of cylindrical panoramic images,position and color calculation of pixels on novel view is formulated.Results of our approach applied to both synthetic and real scene are given at the end of this paper.