A new algorithm of fast SSD calculation for motion estimation
E A Altman, Захаренко Елена Игоревна · Journal of Physics Conference Series · 2018
This paper considers motion estimation algorithms using a Sum of Squared Differences (SSD) metric and making an exhaustive search of motion vectors. A new fast algorithm, based on a new method of fast 2D convolution, is presented. It reduces a number of required arithmetic operations for finding all motion vectors in some area, compared to early published algorithms. The paper gives equations, underlines the presented algorithm, shows the tables comparing a new algorithm to another one in various cases. The new algorithm reduces the computational amount to 3.4%.