二阶广义全变分耦合非局部变换域模型的图像放大

Image enlargement basing on second order total generalized variation coupling to non-local transform domain model

为了提高偏微分方程放大算法对弱边缘和纹理细节的放大效果,采用二阶广义全变分耦合非局部变换域模型,提出了一种图像放大算法模型。非局部变换域模型通过对相似图像块构成的图像组进行三维变换,利用由于图像非局部自相似特性导致的变换系数稀疏特性建模,能够很好利用图像中相似图像块的非局部信息。该算法利用二阶广义全变分模型增强强边缘,非局部变换域模型增强弱边缘和纹理细节,通过变分模型实现两者的耦合,具有良好的放大效果。与其他算法进行仿真实验比较,二阶广义全变分耦合非局部变换域模型在处理强边缘、弱边缘和细节上都取得了较好的放大效果。

In order to improve the zoomed effect of the weak edge and texture of image,the image enlargement model is proposed based on partial differential equation,combining total generalized variation and non-local transform domain model.Using the non-local self-similarity property of the image through the three dimension transform of the group composed of similar image block,the sparse representation model in transform domain utilizes non-local information of the image effectively.Total Generalized Variation enhancing strong edges better and non-local transform domain model enhancing the weak edge and texture details better,the proposed model has better performance.Compared with other algorithms,the second-order generalized total variation coupling to non-local transform domain model achieves better amplification effect in dealing with strong edges,weak edges and details.