基于小波域改进HMT模型的图像恢复算法

针对图像恢复中的边缘模糊问题,提出了一种基于小波域改进隐马尔可夫树( IHMT)模型的图像恢复算法。IHMT模型更多描述了相邻尺度小波系数的互相关性,能准确刻画自然图像小波系数的统计特性。本文从图像恢复的贝叶斯框架出发,将简化的IHMT模型作为图像小波域的先验模型,构造正则化约束进行图像恢复。采用近似等价的方法,将含有混合密度的恢复方程简化为单一密度求解。实验结果表明,该算法能有效再现图像的边缘信息,提高峰值信噪比。

Image Restoration Algorithm Based on Wavelet Domain Improved Hidden Markov Tree Model

An image restoration algorithm based on improved hidden Markov tree(IHMT) model in the wavelet domain is proposed,faced with the edge blurring problem in image recovery.More cross-correlations of wavelet coefficients between two neighboring scales are captured in the IHMT model,which makes the statistical property of the real world images' wavelet coefficients are accurately characterized.In light of Bayesian framework of image restoration method,simplified IHMT model is used as a priori information of images in the wavelet domain,and regularization restriction is made to recover them.An approximate equivalence method is used to simplify the recovery equation with mixture densities into single density.Experimental results show that,the algorithm can effectively retrieve edge information of images,and the peak signal to noise ratio(PSNR) is improved significantly.