A complexity-bounded motion estimation algorithm

Motion estimation full search algorithm in video coding is an operation computationally very intensive. For this reason, in real time or low power application, simplified motion estimation algorithms are very appreciated. In this paper a new motion estimation algorithm of reduced complexity will be presented. It tries to conjugate the extreme reduction of computational load with a good efficiency. It exploits the "past" history of the motion field to predict the current motion field. This approach permits to limit greatly the number of motion vector to test. A following refinement phase gives the final motion vector. The complexity is lower than other algorithms present in literature and is also constant because in the algorithm there is no recursivity. Simulation evaluation show the robustness of the algorithm with any kind of video sequence, an optimal performance with respect to other reduced complexity algorithms and a very reduced loss of efficiency with respect to the full search algorithm