作者: Brendt Wohlberg , Paul Rodriguez
关键词: Mathematics 、 Norm (mathematics) 、 Mathematical optimization 、 Image denoising 、 Fidelity 、 Deconvolution 、 Algorithm 、 Noise reduction 、 Inverse problem 、 Image restoration 、 Minification
摘要: Total variation (TV) regularization has become a popular method for wide variety of image restoration problems, including denoising and deconvolution. A number authors have recently noted the advantages replacing standard lscr2data fidelity term with an lscr1 norm. We propose simple but very flexible solving generalized TV functional that includes both lscr2 -TV problems as special cases. This offers competitive computational performance is comparable to or faster than any other algorithms which we are aware.