Blind Unitary Transform Learning for Inverse Problems in Light-Field Imaging

Cameron J. Blocker, Jeffrey A. Fessler; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 0-0

Light-field cameras have enabled a new class of digital post-processing techniques. Unfortunately, the sampling requirements needed to capture a 4D color light-field directly using a microlens array requires sacrificing spatial resolution and SNR in return for greater angular resolution. Because recovering the true light-field from focal-stack data is an ill-posed inverse problem, we propose using blind unitary transform learning (UTL) as a regularizer. UTL attempts to learn a set of filters that maximize the sparsity of the encoded representation. This paper investigates which dimensions of a light-field are most sparsifiable by UTL and lead to the best reconstruction performance. We apply the UTL regularizer to light-field inpainting and focal stack reconstruction problems and find it improves performance over traditional hand-crafted regularizers.