Efficient representation of spatio-temporal data using cylindrical shearlets

Efficient representations of multivariate functions are critical for the design state-of-the-art methods data restoration and image reconstruction. In this work, we consider representation spatio-temporal such as temporal sequences (videos) 2- 3-dimensional images, where conventional separable usually very inefficient, due to their limitations in handling geometry data. To address challenge, define a class E(A)⊂L2(R4) 4 variables dominated by hypersurface singularities first three coordinates that apply model 4-dimensional corresponding objects. provide an efficient type data, introduce new multiscale directional system based on cylindrical shearlets prove approach achieves superior approximation properties with respect representations. We illustrate advantages our applying discrete implementation challenging problem from dynamic tomography. Numerical results confirm potential novel methods.