Revision history [back]

Use the CUDA implementation of template matching if you have a nvidia video card.

The second (and mode general) solution is to implement your algorithm in OpenCL. However this is quite complicated...