An improved morphological wavelet construction based on image statistical information

According to the correlation and statistical information of neighboring pixels in image, a statistical morphological predict operator is constructed based on the Markov random field theory, and the corresponding 2-D statistical morphological wavelet is also defined. Under the conditional probability density function in MRF model, It can be proved that more detail coefficients approach to zero after the image is decomposed by the statistical morphological wavelet. We compare the 2-D statistical morphological wavelet with four classical morphological wavelets and 5/3, 9/7 lifting wavelet in image lossless coding. Experimental results indicate good performances of the proposed method with small entropy for smooth low complexity image.