基于小波域三状态HMT模型的含噪图像增强

针对含噪图像增强问题，提出一种基于小波域三状态隐马尔可夫树模型的方法,采用三状态的高斯混合模型逼近小波系数的分布,不需要设定精确的阈值,依据期望最大算法训练得到的每个系数所属状态的后验概率,将系数区分为噪声系数、弱边缘系数和强边缘系数,然后通过抑制噪声系数,增强细节特征系数来达到对含噪图像增强的目的,并引入循环平移策略避免人工失真.通过对含噪的标准图像和人脑核磁共振图像进行仿真实验,并与几种经典的图像增强方法作视觉上的对比和定量分析.实验结果表明,本文所提出的方法具有很好的鲁棒性,在突出了图像中更多的细节信息的同时,可以有效抑制噪声.

Noisy Image Enhancement Based on Three-state HMT Model in Wavelet domain

A noisy image enhancement method is proposed based on the three-state hidden Markov tree model in wavelet domain.It is not need to confirm thresholds accurately,the three-state Gaussian mixture model is adopted to estimate the distribution of wavelets coefficients,according to the states posterior probability of each coefficient belongs to achieving by the training of expectation maximization algorithm,coefficients are distinguished into noise,weak edge and strong edge coefficients respectively.Then the enhanced noisy image is obtained by restraining noise coefficients and enhancing detail feature coefficients.Cycle spinning strategy is introduced to avoid visual artifacts.By experimenting on noisy standard images and brain magnetic resonance images,compared with several classical image enhancement methods in visual effects and quantitative analysis,experiments show that the enhancement method proposed bears better robustness,can emerge more detail information and restrain noise effectively at the same time.