A novel parameter estimation algorithm based on RANSAC for precise omnidirectional image unwrapping

Conventional perspective cameras is restricted for certain application in computer vision due to the limited fields of view. Omnidirectional camera, on the other hand, has large fields of view, which is hard for human observation and traditional visual algorithm application. Therefore, approaches to unwrap the distorted image into perspective image are required. As the quality of unwrapping omnidirectional image is heavily depend on the estimation of system parameters. In this work, we propose an improved algorithm to solve the problem of parameter estimation, using RANSAC and omnidirectional vision imaging characteristics. Experiment shows that the algorithm has high accuracy, strong anti-interference characteristics.