Image Edge Detection Based on Gaussian Mixture Model in Nonsubsampled Contourlet Domain

Algorithm 1

Data screening algorithm by Gaussian mixture model.

Input: high-frequency directional sub-bands

() Employ (1)~(3) to calculate the probability of generated by each component

() Employ (4)~(7) to calculate the , ,

() Repeated iteration, until the values of the likelihood function achieves convergence.

() Calculate the probability of generated by each component according to , ,

, . Let pro0 be the probability of generated by the “big” variance of Gaussian

model; let pro1 be the probability of generated by the “small” variance of Gaussian

model.

() If pro0 < pro1, set 0 to

Output: (new high-frequency directional sub-bands )