A fast multi-reference frame motion estimation algorithm

This paper proposes a new fast multi-reference frame motion estimation algorithm for efficient hardware implementation. The proposed algorithm reduces the search areas of motion estimation using a linear relationship between the motion vector and the distances from the current frame to the reference frames. First of all, it executes a full search area motion estimation in reference frames 0 and 1. After that, the search areas of motion estimation in reference frames 2, 3 and 4 are minimized by a linear relationship between the motion vector and distances from the current frame to the reference frames. The proposed algorithm does not use an early termination which gives us a fixed processing time schedule and a small size area of hardware. Simulation results show that PSNR drop and bit-rate increase are lower than those of previous fast multi-reference frame motion estimation algorithms. The number of computations of the new algorithm is reduced by 52.5% and the sequence quality of the simulation results is almost the same as with a full search area motion estimation which uses 5 reference frames.