Real-time recursive motion segmentation of video data on a programmable device

We previously reported on a recursive algorithm enabling real-time object-based motion estimation (OME) of standard definition video on a digital signal processor (DSP). The algorithm approximates the motion of the objects in the image with parametric motion models and creates a segmentation mask by assigning to every block in the image the best matching model. It was found that the calculation of the segmentation mask was most critical for both the real-time behaviour and the quality of the motion vector field. This publication details the motion segmentation module and discusses a number of options which improve the quality/complexity ratio of the motion segmentation.