An Efficent Variable Block Size Selection Scheme for the H.264 Motion Estimation

An Enhanced Lenient Merging Scheme for H.264 Variable Block Size Selection

“…It halves the bitrate necessary for the same quality resulting from any other existing video coding standards [1]. In March 2003, the standard was finalized and submitted for formal approval as H.264/AVC.…”

“…If the 8×8 mode is chosen, then there can be further division either as one 8×8 partition, two 8×4 partitions, two 4×8 partitions or four 4×4 partitions [3][4]. As a result, there are 259 (1 (16×16) mode, 1 (16×8) mode, 1 (8×16) mode, 4 4 (8×8) modes) possible combinations in MB partitioning [1]. In the mode decision step, a specific MB partitioning scheme is chosen based on the possible minimum distortion that can be achieved at a specific bitrate (i.e.…”

“…minimum R-D cost). This can be acquired by using the following formula [2]: Find min (J) where, J = D + R where J is the R-D cost, D is the amount of distortion between the reconstructed and the original MB, R is the number of bits representing the motion vector(s) and is the Lagrange parameter which represents the relative significance of distortion with respect to bitrate [1]. This VBS ME/MD tool enhances the motion tracking capabilities of the encoder by representing active regions with small block size modes and homogenous regions with large block size modes.…”

“…It performs motion estimation for all the defined seven types and then the best partitioning scheme is chosen from all the 259 possible combinations. This dramatically increases the computational requirements to select a suitable partitioning scheme for every certain region within the image frame [1]. A recently proposed merging scheme for efficient VBS ME/MD was introduced in [1] to reduce the computational requirements of the JM model at the expense of minor degradation in quality.…”

“…The algorithm has three different operating points with different computation requirements to meet different application needs. This paper proposes an enhancement to the merging scheme of [1] by modifying the merging rule between a pair of block nodes. The remainder of this paper is organized as follows: In Section 2, a brief description of the scheme in [1] is provided.…”