基于熵变分的图像去噪模型

针对传统Tikhonov正则化模型存在的不足，根据最大后验概率（MAP）和最大熵的理论，提出了一种基于熵变分的图像去噪模型。该模型利用图像像素点的梯度信息自适应的对带噪图像进行各向异性滤波处理，在去除噪声的同时有效保留了图像的边缘细节。采用变分法推导出了该模型对应的偏微分方程，最后结合梯度加权最速下降法和半点格式的数值迭代算法对方程进行求解。实验结果表明，该模型去噪后的图像比Tikhonov正则化模型具有更好的客观评价指标和主观视觉效果。

The Image Denoising Model Based on Entropy Variation

An image denoising model based on entropy variation is proposed from the viewpoint of maximum a posteriori probabilities(MAP) and maximum entropy theory.Compared with conventional Tikhonov regularization method,the proposed model can adaptively deal with the noise image with the gradient information of each pixel,which is used as an anisotropic filter.It not only suppresses noises very effectively but also preserves image edge details very well.Then the partial differential equation of the proposed model is...