Colloquium| Institute of Mathematical Sciences
Time：14:00-15:00, October 10 Thursday
Location：Room S408, IMS
Speaker: Jae Kyu Choi, School of Mathematical Sciences, Tongji University
Abstract: Image restoration aims at recovering an image of high-quality from a given degraded measurement. Since it requires solving an ill-posed linear inverse problem in general, the sparse regularization based approaches are most widely used to regularize designed smooth image components while preserving key features or image singularities such as edges, ridges, corners, etc. However, since the degraded measurements are available in general, it leaves an ambiguity in estimating the image singularities as accuarately as possible. In this talk, we introduce three related recent works. One is the edge driven wavelet frame based model which is designed to restore piecewise smooth images via different strength of regularization on smooth and singular image regions and singularities. Second, we introduce a joint sparsity based joint reconstruction model which aims to borrow the complementary information of each modality to the other. Finally, we introduce the continuous domain image restoration model which aims to resolve the basis mismatch between the original singularity in continuum and the discrete image domain.