Hardware Design for the New Galaxy Random Search Motion Estimation Algorithm Focusing in HD Videos

This paper presents a new algorithm for motion estimation in high definition videos and it also proposes a hardware architecture to perform it. The algorithm uses a random search in the candidate blocks as strategy to avoid local minima falls and to increase the motion estimation quality . This algorithm is called Galaxy Random Search (GRS). Software evaluation shows that the GRS algorithm can achieve a PSNR gain of 1.82 dB in comparison to Diamond Search algorithm, when 1080p video sequences are considered. The proposed architecture was designed focusing on real time processing for HD 1080p videos. The preliminary results show that the proposed architecture might be able to process 1080p videos in real time (at 30 frames per second) with similar hardware costs than a Diamond Search architecture.