I am currently using the opencv library for implementing stereo vision on the raspberry pi. A colleague of mine is using an odroid sbc supporting opencl for parallel processing of matrices.
Is opencl supported in Raspberry Pi for stereo vision applications?
The answer isn't totally straightforward, but I suspect that for your purposes the answer is 'no' at present.
Portable OpenCL on the CPU (I think this will be too slow for your application) does seem to work: http://parallelplusplus.blogspot.co.uk/2015/10/opencl-on-raspberry-pi-2.html?m=1
But folks seem to be making some progress at getting it running on the GPU: http://petewarden.com/2014/08/07/how-to-optimize-raspberry-pi-code-using-its-gpu/ https://m.reddit.com/r/raspberry_pi/comments/3fi2of/is_there_an_opencl_implementation_for_the_pi/
But it could be a while before we see the fruits of their labours.
