https://hgpu.org/?p=1276
Parallel implementation of a spatio-temporal visual saliency model