Edge Deection by Weak Regularization
弱正则化边缘检测

A new technique of edge detection which called weak regularization approach is presented. The technique consists of two stages. The first is a weakly regularized reconstruction (WRR) of image functions, the second is an edge finding (EF)procedure. The WRR is characterized by a controlled-smoothness stabilizer in the reconstruction functional by means of which smoothness constraint on the reconstructed image is adjusted (released or tightened) with the possibility of occurrence of edges in the underlying image. What is new in the EF is an edge pruning technique that yields optimal pruning by minimizing a cost function for the evaluation of local geometrical structure of edges. Experiment results that demonstrate the performance of the technique are given.