Image demosaicing and rectification are key tasks that are frequently used in many computer vision systems. To date, however, their implementations have been plagued with large memory requirements and inconvenient dataflow, making it difficult to scale them to real-time, high resolution settings.
