Abstract: Using the Local Contextual Hidden Markov Model（LCHMM） in uniform discrete curvelet domain, an image denoising algorithm is proposed. After introducing the characteristics of the new transform, the statistical distribution rules of it are analyzed, which shows that the hidden markov model is suited to model the new transform’s coefficients. The estimative coefficients of denoised image can be abtained by the model’s parameters, which are captured through expectation maximization training method. The proposed algorithm is applied to denoising the optical image and high resolution SAR image respectively. Compared with the LCHMMs in wavelet and contourlet domain, the experimental results show that the proposed algorithm can reduce noise effectively with well edge-preserving ability.

Key words: image denoising, uniform discrete curvelet transform, local contextual, hidden markov model

摘要： 提出了一种在均匀离散曲波域中利用局部上下文隐马尔可夫模型进行建模的图像降噪算法。介绍均匀离散曲波变换的特点，分析其系数的统计分布规律，表明适合用隐马尔可夫模型对其进行建模。通过期望最大化训练获取模型的参数，利用参数得到降噪图像的系数估计。分别对光学图像和高分辨率的SAR图像进行了降噪实验，与小波域、轮廓波域的局部上下文隐马尔可夫模型等降噪方法进行比较，结果表明，提出的算法能够有效地去除噪声，具有较强的边缘保持能力。

关键词: 图像降噪, 均匀离散曲波变换, 局部上下文, 隐马尔可夫模型