Deeply Learned Spectral Total Variation Decomposition

In that, we show the optimality of our network in terms of loss functionals and highlight the generalisibility further. Moreover, we consider visual and quantitative results for images of ellipses. TVspecNET achieves high quantitative performance measures: SSIM is 0.9658, PSNR is 30.609 We show the visual results for an example ellipse image in Figure 3. Documentation can be found at https://odlgroup.github.io/odl/ All bands have the same scaling.