Uniform optical flow model by foveation effect for virtual cylindrical screen algorithm

This paper focuses on extending the Virtual Cylindrical Screen (VCS) algorithm for detecting stationary objects inside a VCS, when a sensor moves forward along the optical axis. Effect by foveation in use of the VCS algorithm are simulated and discussed by comparing a pinhole camera (PHC) model with 2 foveation models, that is, a sine curve (SC) model and Advanced Wide Angle Foveated (AdWAF) model. The foveation models reduce the optical flow and also its variation. Correction weight function (CWF) is defined for estimating error of the optical flow computed in each image model. Moreover, Uniform Optical Flow (UOF) model is designed as a foveation model for the VCS algorithm, and its properties are discussed. Accuracy of the optical flow corresponding to the VCS is simulated using CWF. AdWAF model has advantages of accuracy comparing with not only the PHC model but also the SC model when the vΔt /R is less than 0.14. Experiments using the Lucas-Kanade method are demonstrated for the images by UOF model.