Abstract: It is important to develop an effective image restoration method which can preserve the small scale image structure such as texture.Since that the natural images always contain various morphological components such as cartoon（piecewise smooth，edge），texture etc，and it is difficult to find a single effective regularization term to constrain the whole image，hence an optimization model is proposed which contains regularization in terms of morphological components.The alternating minimization scheme is adopted to solve the relevant multi-variable optimization problem，and the TwIST algorithm is applied to solve the relevant sub-problem.Compared with the methods of min-TV and min-l1，the proposed method can preserve well the whole visual quality and the small scale image structure such as texture.

Key words: image restoration, morphological components, alternating minimization

摘要： 如何设计能够保持图像纹理等小尺度结构特征的图像恢复方法是目前该领域有待解决的难点问题。由于自然图像往往包含卡通（平滑、边缘）、纹理等多种形态结构成分，很难找到单一有效的正则项对整幅图像进行约束。因此将各形态成分分开处理，建立多形态成分正则化的图像恢复最优化模型。采用交替最小化策略，对相应的多变量优化问题进行数值求解，每一子问题采用TwIST算法进行快速求解。仿真实验结果显示与min-TV和min-l1方法相比，形态成分正则化方法可以较好地保持恢复图像的整体视觉效果及纹理等小尺度结构特征。

关键词: 图像恢复, 形态成分, 交替最小化