Sub-pixel location of edges with non-uniform blurring: a finite closed-form approach

摘要

Traditional sub-pixel edge detectors model edges as step changes in intensity. However, this assumption is unrealistic in natural images, where edges have varying degrees of blur. This paper generalizes the edge model to one containing a local blurring factor, and gives finite closed-form solutions for edge location with sub-pixel accuracy. Solutions are based on the spatial moments up to the second-order, and are theoretically precise within the finite size of a windowed area. Moreover, the integral nature of low order spatial moments provide the algorithm some resilience against additive white noise. Experiments on both synthetic and real images show that the accuracy of the proposed method is comparable to the more computationally demanding nonlinear optimization method. Absolute error of less than 0.02 pixel is demonstrated on one image of a grid pattern.