Hey,
I would like to use the gpu-implementation of Optical Flow in OpenCV. Therefore I compiled OpenCV (2.4.9) with CUDA (6.0) support. The disparity between my two pictures is VERY large. To get a good Optical Flow estimation, I would like to initialise the u-component with a already computed flow field and not with zeroes. Does anyone of you have an idea where I have to change the code?
Thanks a lot!