SPRUIL1D May 2019 – December 2024 DRA829J , DRA829J-Q1 , DRA829V , DRA829V-Q1 , TDA4VM , TDA4VM-Q1
The DMPAC implements Texas Instruments’ proprietary algorithm to find the optical flow vector map between the input image pair. The optical flow estimation process implemented in the DMPAC has been formulated by combining pyramidal block matching (PBM) method with differential technique based optical flow method proposed by Lucas-Kanade(LK). In this formulation the PBM method performs optical flow estimation with integer pixel resolution and followed by an iteration of the LK method used to refine the obtained estimate and obtain optical flow estimate with fractional pixel precision. The PBM method starts the optical flow estimation at the highest pyramid level using spatial predictors and step search method to obtain accurate motion estimation. These optical flow estimates are then appropriately scaled up and refined (again using spatial predictors and step search method) at each of the lower pyramid levels sequentially. At base pyramid level the predictor configuration can be altered to include temporal predictors or auxiliary predictors, however step search remains the same. At base pyramid level, once PBM process completes, an iteration of LK step is used to refine the obtained flow estimates. During predictor evaluation and step search the PBM method uses binary census transform for pixel description and hamming distance as the optimization function. The 2D post filtering of the optical flow estimates can also be performed at output of each of the pyramid levels including base level. The algorithm has been optimized to fit into real time, low power processing requirements with highly precise and accurate estimates.
This section outlines the overall functionality of the Dense Optical Flow (DOF) core.