Mask-based non-maximal suppression with iterative pruning for low-complexity corner detection

摘要

Adaptive and low-complexity corner detection is essential for the realization of modern real-time video processing applications on low-resource embedded platforms. In our earlier work, we have proposed an efficient iterative pruning strategy for widely used Shi-Tomasi and Harris corner detectors. In this paper we propose a mask-based strategy for efficient nonmaximal suppression. In comparison to the conventional nonmaximal suppression, applying the proposed strategy results in higher efficiency in computations for the iterative pruning based Shi-Tomasi/Harris corner detectors. On the embedded NiOS-2 platform, we achieve a speedup in execution time of 14-42% for Shi-Tomasi and 11-47% for Harris corner detectors.