摘要 |
A method of transforming a stream of wide-angle video frames, the method comprising mapping each video frame onto a spherical mesh selecting one or more features present in a number of the video frames for each frame, generating a energy equation for each of a plurality of image transformation criteria and minimising the energy equations to define an image transformation and applying the image transformation to the spherical mesh for said frame, wherein one of the transformation criteria comprises a measure of temporal coherency between successive frames, whereby the minimisation of each energy equation for each frame includes minimising the change in transformation applied to the selected features between successive frames. |