Omnidirectional Image
Omnidirectional images (ODIs), capturing a 360-degree field of view, are increasingly important for applications like virtual reality and robotics. Current research focuses on overcoming challenges posed by the inherent distortions of these images, employing techniques like generative adversarial networks (GANs), transformers, and Siamese neural networks to improve image synthesis, super-resolution, depth estimation, and quality assessment. These advancements are crucial for enhancing the realism and usability of ODI-based systems, impacting fields ranging from immersive media to autonomous navigation.
Papers
Applications of Deep Learning for Top-View Omnidirectional Imaging: A Survey
Jingrui Yu, Ana Cecilia Perez Grassi, Gangolf Hirtz
Human Pose Estimation in Monocular Omnidirectional Top-View Images
Jingrui Yu, Tobias Scheck, Roman Seidel, Yukti Adya, Dipankar Nandi, Gangolf Hirtz
360$^\circ$ High-Resolution Depth Estimation via Uncertainty-aware Structural Knowledge Transfer
Zidong Cao, Hao Ai, Athanasios V. Vasilakos, Lin Wang