分数阶整体变分泊松去噪模型的分裂Bregman方法

Split Bregman method for a fractional-order total variation poisson denoising model

整体变分正则化模型在去噪时能较好的保边,但是不能很好地保持纹理细节和图像对比度,且去噪后图像在平坦区域会出现阶梯效应。为此,提出了一种新的分数阶整体变分(Fractional-order Total Variation,FOTV)泊松去噪模型,以及分裂Bregman方法求解。通过与求解整体变分泊松去噪模型的分裂Bregman方法比较,数值实验表明,所提出的模型进行泊松去噪时视觉效果较好,且去噪图像具有较高峰值信噪比。

The total variation regularization model performs very well for preserving edges while removing noise.However,small details such as textures are often filtered out with noise.It can not preserve image contrasts well and often leads to staircase effects in flat region.Therefore,in this paper,we propose a new fractional-order total variation model for poisson denoising.Furthermore,we propose a split Bregman method for solving this model.Compared with the split Bregman method for solving the total variation poisson denoising model,numerical results show that our method performs better in terms of the visual quality and PSNR value.