光谱成像技术研究室知识库

A panorama can reflect the surrounding scenery because it is an image with a wide angle of view. It can be applied in virtual reality, smart homes and other fields as well. A multi-directional reconstruction algorithm for panoramic camera is proposed in this paper according to the imaging principle of dome camera, as the distortion inevitably exists in the captured panorama. First, parameters of a panoramic image are calculated. Then, a weighting operator with location information is introduced to solve the problem of rough edges by taking full advantage of pixels. Six directions of the mapping model are built, which include up, down, left, right, front and back, according to the correspondence between cylinder and spherical coordinates. Finally, multi-directional image reconstruction can be realized. Various experiments are performed in panoramas (1024x1024) with 30 different shooting scenes. Results show that the azimuth image can be reconstructed quickly and accurately. The fuzzy edge can be alleviated effectively. The rate of pixel utilization can reach 84%, and it is 33% higher than the direct mapping algorithm. Large scale distortion is also further studied.