加权各向异性总变差最小化器的不连续性研究
Discontinuities of the minimizers of the weighted or anisotropic total variation for image reconstruction
AI总结:
本文研究了加权各向异性总变差最小化器的正则性,证明了通过适当选择权重和各向异性可生成原图像不含的不连续性,并通过数值模拟验证了理论结果。
摘要:
本文研究了加权各向异性总变差与一般保真项最小化器的正则性。我们推广了最近关于图像去噪问题最小化器不连续性的结果。特别地,我们证明了通过适当选择权重和各向异性,可以生成原图像中不存在的不连续性。我们还观察到最小化器不连续处的跳跃性质有所减弱。为了证明这些结果,我们使用了一些关于极小曲面的正则性定理,并将其适应到我们的设定中。我们还通过多个数值模拟展示了我们的理论结果。
英文摘要:
The study of the regularity of the minimizer of the weighted anisotropic total variation with a general fidelity term is at the heart of this paper. We generalized some recent results on the inclusion of the discontinuities of the minimizer of the image denoising problem. In particular, we proved that for well-chosen weights and anisotropies, it is actually possible to create discontinuities that were not contained in the original image. We also observed a reduced jump property at the discontinuities of the minimizer. To prove these results we used some regularity theorems for minimal surfaces that we had to adapt to our setting. We also illustrated our theoretical results with several numerical simulations.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。