Refined adaptive search range for predictive line search motion estimation algorithm
Hung-Ming Chen, Po-Hung Chen, Feng-Hsu Tsai, Mon-Chau Shie · 2010
PLS algorithm has been proved successfully in block motion estimation on media processors due to low control overhead and high data reusability. These characteristics causes PLS also suitable for hardware VLSI implementation. However, the line search pattern of PLS results in high computational complexity since it searches all points in a horizontal line of the predefined search range. To improve this issue, a refined adaptive search range (RASR) technique based on the motion activities is proposed to define the horizontal and vertical search ranges separately, which can effectively reduces redundant search points in the line search pattern of PLS. Experimental results show that RASR significantly improves 63% computational speed averagely over the original PLS for block motion estimation.