Improved CNN algorithm for H.264 motion estimation partitions

To take full advantage of the motion estimation of the new video coding standard H.264, low-power video encoders will need specific hardware accelerators. Presented here is an improved partitioning method to decrease the computational load of variable block size motion estimation. The two-step partitioning method improves on a previous three-step method. The method is derived from a cellular nonlinear network (CNN) segmentation algorithm and, along with the partition, indicates early termination of motion estimation and the skip modes of H.264. The algorithm achieves better rate-distortion performance when compared to motion estimation with only 16×16 sized blocks and, for low bit rates, equivalent performance when compared to Lagrange optimization.