中文摘要：

从贝叶斯估计出发，构造了一种新的变分模型，用于复原被泊松噪声污染的模糊图像。首先讨论了模型正则化项中具有边缘保持能力的函数选取以及模型求解的相关问题，然后将变分模型的求解转化为可快速求解的非线性扩散方程，给出了正则化参数选取的初步空间自适应方法，可以区分平滑区域和图像边缘自适应的调节参数。实验结果表明，本文方法的复原效果整体上优于传统的迭代正则化方法，复原图像的边缘得到了有效的保护，泊松噪声的抑制效果明显，复原图像提高的改进信噪比（ISNR）要比迭代正则化方法平均提高1dB以上。

英文摘要：

The restoration of blurred image with Poisson noise was studied. According to the statistical maximum a posteriori MAP estimation of the original image,we built a new criterion to measure the fidelity of the estimated image to the original image corrupted by Poisson noise, Because of the ill-posed nature of the image restoration problem,we construct a new variational model with a regularization term. The choice of the edge-preserving regularization function.is addressed. To solve the variational model,we transform it to be a nonlinear diffusion equation. An adaptive regularization parameter, which can change its value from a smooth area to an edge area, is proposed. Numerical experiments demonstrate that the proposed method results in high restoration performance. The new model can preserve edges and reduce the Poisson noise effectively. The improved signal to noise ratio（ISNR） is the new model is about 1 dB higher than the traditional iterative regularization method.