Abstract:

In order to effectively eliminate the noises existing in boiler coal combustion flame images that are unfa-vorable for the subsequent image feature extraction and temperature reconstruction,an image denoising methodbased on the HMT (Hidden Markov Tree) model in dual- tree complex wavelet domain is proposed.In this method,first,a dual- tree complex wavelet transform is performed for noisy flame image.Next,the real part and the imagi-nary part of the dual- tree complex wavelet coefficients are respectively modeled according to the HMT model.Then,the model parameters are estimated by using the expectation maximization algorithm,and the noiseless dual- treecomplex wavelet coefficients are estimated according to the Bayes minimum mean square error (MMSE) criterion.Finally,an inverse dual- tree complex wavelet transform is conducted to obtain denoised flame images.Experimentalresults show that the proposed method is superior to the wavelet VisuShrink threshold method and the method basedon HMT model in wavelet or Contourlet domain because it helps to reduce the noise more effectively and achievehigher peak signal- to- noise ratio.

Key words: image processing, image denoising, boiler coal combustion, flame image, dual- tree complex wavelettransform, hidden Markov tree model, expectation maximization algorithm, Bayes estimation