Probabilistic-based Feature Embedding of 4-D Light Fields for Compressive Imaging and Denoising
This addresses the problem of high-dimensional light field processing for computational imaging applications, representing an incremental improvement with a novel method for a known bottleneck.
The paper tackles the challenge of efficient feature embedding for 4-D light fields, which impacts downstream tasks like compressive imaging and denoising, by proposing a probabilistic-based feature embedding method that assembles low-dimensional convolution patterns in probability space; experiments show significant superiority over state-of-the-art methods on real-world and synthetic data.
The high-dimensional nature of the 4-D light field (LF) poses great challenges in achieving efficient and effective feature embedding, that severely impacts the performance of downstream tasks. To tackle this crucial issue, in contrast to existing methods with empirically-designed architectures, we propose a probabilistic-based feature embedding (PFE), which learns a feature embedding architecture by assembling various low-dimensional convolution patterns in a probability space for fully capturing spatial-angular information. Building upon the proposed PFE, we then leverage the intrinsic linear imaging model of the coded aperture camera to construct a cycle-consistent 4-D LF reconstruction network from coded measurements. Moreover, we incorporate PFE into an iterative optimization framework for 4-D LF denoising. Our extensive experiments demonstrate the significant superiority of our methods on both real-world and synthetic 4-D LF images, both quantitatively and qualitatively, when compared with state-of-the-art methods. The source code will be publicly available at https://github.com/lyuxianqiang/LFCA-CR-NET.