An efficient approach to the identification of characteristic intensity variations

摘要

The approach of Rohr1,2 for identifying characteristic intensity variations in real images uses a parametric model consisting of the superposition of wedge-shaped structures smoothed by a Gaussian filter. To reduce the computation time in this contribution we derive an approximation of the original model which merely consists of polynomials, and apply the minimization method of Levenberg-Marquardt3. Using the approximate model, the identification results remain nearly the same while the computation costs can considerably be reduced.