A Novel Parallel Motion Estimation Algorithm Based on Particle Swarm Optimization Manal K. Jalloul and Mohamad Adnan Al-Alaoui ECE Department, American University of Beirut, Beirut, Lebanon Abstract— Motion estimation is a common tool used in all video coding standards. Fast and accurate algorithms are needed to target the real-time processing requirements of emerging applications. Many fast-search block motion estimation algorithms have been developed to reduce the computational cost required by the full-search algorithm. These techniques however often converge to a local minimum, which makes them subject to noise and matching errors. In the literature, several schemes were proposed to employ strategies of Particle Swarm Optimization (PSO) in the problem of motion estimation since PSO promises to alleviate the problem of being trapped in local minima. The existing schemes, however, still don’t achieve the necessary improvement in terms of accuracy or speedup as compared to the existing fast searching methods. In this paper, we propose a novel fast and accurate block motion estimation scheme based on an improved parallel Particle Swarm Optimization algorithm. Unlike existing motion estimation algorithms which operate on blocks of the frame serially following the raster order, the proposed algorithm achieves parallelism since it performs motion estimation for all blocks of the frame in parallel. Simulation results showed that the proposed scheme could provide a higher accuracy and a remarkable speedup as compared to the well- known fast searching techniques and to a recent PSO-based motion estimation algorithm. I. INTRODUCTION With the proliferation of several cutting-edge real time multimedia applications such as video streaming, video conferencing, and high definition TV (HDTV), there is a growing need to accelerate the motion estimation process to meet the real-time processing requirements. Block-Matching Motion Estimation (BMME) with Full Search (FS) algorithm is the main computational burden in the video encoding process due to exhaustively search all possible blocks within the search window. In the literature, many approaches were researched to reduce the computational cost of the Exhaustive FS method. In the past years, these algorithms included three-step search (3SS) [1], four-step search (4SS) [2] which can be generalized to N- step search (NSS), the diamond search (DS) methods [3], the cross-diamond search (CDS) method [4], and the Hexagon- based search [5]. In each of these fast search methods, a different search pattern is employed to reduce the number of search points. These algorithms reduce the computational complexity with negligible loss of image quality only when the motions matched the pattern well; otherwise, the image quality will decrease. Block matching motion estimation can be formulated into an optimization problem where one searches for the optimal matching block within a search region which minimizes a certain cost. The above fast block matching methods suffer from poor accuracy since they dictate that only a very small fraction of the entire set of candidate blocks be examined, thereby making the search susceptible to being trapped into local optima on the error surface. In order to escape from the problem of local minima; several approaches were recently presented in the literature that use modern optimization algorithms to solve the problem of motion estimation. In [6- 8], the Genetic Algorithm (GA) and Simulated Annealing (SA) have been considered for motion estimation (ME). The proposed algorithms, however, tend to be complex and suffer from a high computational burden. In addition to GA and SA, there have been some attempts in the literature to apply Particle Swarm Optimization (PSO) to solve the problem of ME [9-15]. The PSO-based motion estimation methods introduced in [9-13] either have higher computational complexity [9] or have lower estimation accuracy [10- 12] than several existing fast search methods, such as the 3SS and the DS method. In [14], authors proposed a new block matching algorithm based on a set of strategies adapted from the standard particle swarm optimization approach. This is done by employing several strategies which include a deterministic pattern for the initial positions of the particles, particle history preservation and exploitation, and adaptive stopping criteria for the PSO iterative search. These strategies allowed to speed up the motion estimation process and gave high motion estimation accuracy as compared to 4SS, DS, and CDS for sequences with high motion content. As for sequences with medium or low motion, the algorithm in [14] still falls behind the other fast algorithms. These algorithms try to improve the speed of convergence of the PSO iterations by choosing, as initial positions of the particles, the motion vectors (MVs) of the adjacent causal blocks in the frame as well as the zero (0, 0) MV. The PSO iterations, however, can achieve faster convergence if we exploit the temporal correlation with the collocated block in the adjacent frame as well. In [15], a new variant of parallel particle swarm optimization (PPSO) known as small population-based modified PPSO (SPMPPSO) is proposed for fast motion estimation. The proposed algorithm in [15] achieves parallelism at the particle level, where the particles of the swarm evaluate the fitness function concurrently. Nevertheless, the algorithm presented in [15], as well as all the other PSO-based ME algorithms in the literature, operate serially on the blocks of a given frame following the raster order. Thus, if we can device a ME algorithm which can operate in parallel on all blocks of the frame, then the speed of