Statistical Dependency of Image Wavelet Coefficients: Full Bayesian Model for Neural Networks

  • Authors:

  • Xingming Long;Jing Zhou

  • Affiliations:

  • Department of Physics, Chongqing Normal University, Chongqing, China 400047;College of Electrical Engineering, Chongqing University, Email: lennydragon@163.com, Chongqing, China 400044

  • Venue:
  • ISNN '09 Proceedings of the 6th International Symposium on Neural Networks on Advances in Neural Networks
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

A novel method based Full Bayesian Model for Neural Network (FBMNN) to study the statistical dependency of wavelet coefficients is presented. To overcome the ignorance of the relationship between wavelet coefficients, we introduce the FBMNN to model joint probability density distribution (JPDF) of Child and Parent wavelet coefficients. According to the characteristics of the suggested FBMNN-JPDF model, its parameters are estimated by reversible jump MCMC (rjMCMC) algorithm. Finally, a practical application on denoising image by using the FBMNN-JPDF model is demonstrated and the result shows that the suggested method can express wavelet coefficients dependency efficiently.