context-aware encoding

Hi, thanks for the release. I'm currently working on context-aware video encoding which aims at setting different QPs for different area within frame to be encoded according to grayscale saliency map where white areas indicate ROI areas. Now I'm struggling at how to implement context-aware encoding in V4L2. Any advices would be very appritiated.