An Auto-adaptive Threshold Pre-detection SUSAN Corner Detection Algorithm

Corner is a significant geometrical characteristic in digital image processing. The accuracy in corner detection has much meaning in image processing and measurement. This article aims to propose an improved threshold determination method for the Smallest Univalue Segment Assimilating Nucleus corner detection algorism. This improved algorism calculates the threshold values separately for each pixel to make corner detection even under different contrast gradients perform normally. Based on the classical SUSAN algorism, corner pre-detection is used to eliminate pseudo corners and reduce calculation amount and thus improve the algorism speed. Realistic experiments prove this method practical.