Fast descriptor extraction method for a SURF-based interest point

Abstract

Speeded up robust features (SURF) is one of the most popular feature extraction algorithms used for detecting, recognising, and tracking objects in various applications. However, it is not suitable for real-time implementation because its operating speed does not satisfy real-time constraints. In this Letter, operating speed is highly dependent on the amount of computations required for the Haar wavelet response (HWR) step, and present a fast descriptor extraction method that eliminates repeated operations in the HWR step is shown. Experimental results show that the proposed method (PM) can achieve operation time-savings of approximately 27% without using additional resources, while the descriptor extraction performance of the PM is exactly identical to that of the original reference software.