Gap Statistic Based Multi-scale Images Edge Detection Method

A new concept called image distribution gap is defined and analyzed in this paper using the difference of gray-level sample distribution, which is originated from the “Gap Statistic” proposed by Tibshirani, et al. After discussing the detection for step and roof edges in random respectively, a multi-scale image edge detection random model and the corresponding algorithm are proposed. Experimental results demonstrated the good anti-noise and multi-scale properties.