Efficient search points reduction and ranking of Motion Estimation algorithms in video coding

In this paper, the performance of commonly used Motion Estimation (ME) algorithms are analyzed in terms of search point reduction and ranking for hardware implementation. Since number of searching point is not a direct measure of hardware performance, we first define, a cost function of accessing data in memory. Then, we make adjustments on the existing ME algorithms to minimize the problem of memory access by reordering the sequence of searching points, where the number of searching points and MAE/PSNR remains the same as the original algorithms. Finally, we calculate the memory cost improvement of each algorithm. Using the present analysis we see that the cost function of accessing data in memory can be saved from 68% to 80% and the number of search point can be saved from 87% to 94%. The experimental results showed that the ranking based on memory access cost for small motion video is SCDS > BBGDS > CDS > DS > 4SS > 3SS > FS and for high motion video the ranking is memory access cost is BBGDS > CDS > SCDS > 4SS > DS> 3SS > FS with lower computational complexity and similar video quality is maintained.