中文摘要：

图像去噪的难点是如何区分噪声和边缘，并在去噪的同时保持边缘信息。通过分析基于变分形式、散度形式和方向拉普拉斯格式的去噪模型的结构与联系，提出了一种基于方向扩散的彩色图像去噪模型。模型利用结构张量的两个特征值的不同组合作为沿边缘和垂直边缘方向的边缘度量，并用其作为自变量设计出一类单调递减函数作为扩散方程的扩散系数。通过实验获得模型最优参数，理论分析和实验表明，新模型具有更强的去除噪声和保留边缘的能力，并且模型参数具有稳定性。

英文摘要：

The most important and difficult issue of image denoising is to distinguish noise and edges, so that denoising process can eliminate noise while preserving edges. The structures and connections of variational methods, divergence formulations and oriented Laplacian based denoising models are analyzed and a directed diffusion based colored image denoising model is pro- posed. The model uses different combinations of the eigenvalues of the structure-tensor as the edge measurements of along edge direction and isophote direction. The diffusion coefficients, which are monotonically decreasing functions, impose these edge measurements as arguments. The model parameters are estimated by experiments. Theoretical analysis and experiments show that the proposed denoising model has better performance of eliminating noise and preserving edges, and the model parameters are stable with resoect to different images.