摘要
全变分(Total Variation,TV)变换作为一种常用的稀疏变换模型,因其在保持图像边缘信息方面具有明显的优势,已经被应用到图像去噪问题中。然而,它通常会产生阶梯伪影。为了克服这个缺点,在该文中,我们引入交叠组合稀疏全变分(Overlapping Group Sparsity Total Variation,OGSTV)代替传统TV变换模型。为了求解该OGSTV去噪模型,我们提出一种基于快速傅里叶变换(Fast Fourier Transform,FFT)和split Bregman算法的快速OGSTV去噪方法。实验结果表明,引入快速傅里叶变换理论后,图像去噪时间明显减少;与其他已有比较好的算法相比,可以获得更好的图像质量,阶梯效应明显改善。
The total variation(TV)transform is always used as a sparse representation and it has been applied to image denoising problem.Although TV model has obvious advantage in preserving image edges,it is easy to introduce some undesired staircase artifacts.To overcome the drawback mentioned above,an overlapping group sparsity total variation(OGSTV)model is proposed for image denoising instead of TV model.By introducing fast Fourier transform and split Bregman algorithm,a fast algorithm is proposed to solve the OGSTV model.The experimental results demonstrate that,after introducing FFT,the denoising time is reduced obviously.Compared with the other state-of-the-art algorithms,our proposed method can get better image quality.After introducing OGSTV,the staircase artifacts can be eliminated evidently.
作者
林志斌
黄智全
颜林明
Lin Zhi-bin;Huang Zhi-quan;Yan Lin-ming(Xiamen intretech Inc.,Fujian Xiamen 361027)
出处
《电子质量》
2020年第10期79-84,共6页
Electronics Quality
关键词
全变分
图像去噪
快速傅里叶变换
交叠组合稀疏全变分
total variation
image denoising
fast Fourier transform
overlapping group sparsity total variation
作者简介
林志斌(1996-),男,算法工程师,硕士研究生,从事计算机视觉(深度学习方向)工作。
