Fast search algorithm for video coding
Majid Jalalpour, Ali Aghagolzadeh · 2015
Block-matching is the most popular technique for motion estimation /compensation in video coding standards. Full search (FS) is the most accurate and simple technique in block-matching algorithms (BMAs) however it is impractical because of a very laborious computations. Some efficient fast algorithms were proposed to decrease the amount of computation in comparison with FS algorithm. In a fast search algorithm, the aim is to limit the search points while the result must be as close as possible to FS result. The proposed algorithm in this paper intends to reduce the computational cost by targeting most possible locations that may lead to achieve minimum block distortion (MBD). The simulation results show that the proposed algorithm decrease the number of calculations (up to 24%) in comparison with Diamond search while it achieves even higher PSNR for most of standard video sequences.