A Feature-Preserving Adaptive Smoothing Method for Early Vision

A novel adaptive smoothing method is proposed for noise removal and feature preservation. Unlike the previous methods, we adopt two proposed measures to detect discontinuities in an image. Spatial variance is employed to form a measure for detecting contextual discontinuities that can be used for feature preservation and control of the smoothing speed, and another measure is proposed to detect variable local discontinuities during smoothing. As a result, the two discontinuity measures are jointly used in the proposed feature-preserving adaptive smoothing scheme. Due to the use of the contextual discontinuity, our smoothing scheme is insensitive to termination times, and the resulting images in a wide range of iterations are applicable to achieve the identical results in various early vision tasks. Relations between recent adaptive smoothing algorithms and ours are discussed. Simulation results show that our algorithm yields favorable smoothing results for various kinds of images in comparison with recent adaptive smoothing algorithms. Moreover, results of applying our smoothing algorithm to various early vision tasks are reported.