A Modified Directional Gradient Descent Searches for Fast Block Motion Estimation

Abstract

To enhance the DGDS algorithm, this paper presents an efficient algorithm named modified directional gradient descent searches (MDGDS). By analyzing the statistical best MV distribution on some well-known test video sequences, a new search pattern is derived with three search rounds including the window center area, cross area and diagonal area based on stationery and quasi-stationery characteristics. Early search termination operations are then applied to the new search pattern before performing DGDS to eliminate unnecessary computations. The enhanced MDGDS algorithm develops a content adaptive technique based on the previously coded frame for early termination thresholds selection to avoid being trapping in local minima. The simulation results show the proposed algorithm provides significant improvement in reducing the motion estimation by about 19% of the average search points saving compared to the fast DGDS algorithm according to different types of sequences, while maintaining a similar bit-rate without losing picture quality.