Fast Omni-Directional Image Super-Resolution: Adapting the Implicit Image Function with Pixel and Semantic-Wise Spherical Geometric Priors

2025-02-09

AI Summary

To address the non-uniform sampling problem in omnidirectional image super-resolution (ODI-SR) under equirectangular projection (ERP), caused by spherical geometric distortion, this paper proposes a lightweight, low-overhead method supporting arbitrary-scale upsampling. Our core contribution is the first adaptation of implicit image representations to the spherical domain, introducing parameter-free pixel-level and semantic-level spherical distortion mappings, coupled with a geodesically aligned resampling strategy to explicitly embed spherical geometric priors. Complemented by a lightweight affine feature transformation, our approach avoids computationally expensive spherical convolutions or multi-face re-projection. Experiments demonstrate state-of-the-art performance across multiple metrics, 3.2× faster inference speed, and continuous, arbitrary-scale reconstruction capability. Ablation studies confirm both the effectiveness and necessity of each component.

Technology Category

• Computer Vision: Representation Learning for Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Learning with Manifolds

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

In the context of Omni-Directional Image (ODI) Super-Resolution (SR), the unique challenge arises from the non-uniform oversampling characteristics caused by EquiRectangular Projection (ERP). Considerable efforts in designing complex spherical convolutions or polyhedron reprojection offer significant performance improvements but at the expense of cumbersome processing procedures and slower inference speeds. Under these circumstances, this paper proposes a new ODI-SR model characterized by its capacity to perform Fast and Arbitrary-scale ODI-SR processes, denoted as FAOR. The key innovation lies in adapting the implicit image function from the planar image domain to the ERP image domain by incorporating spherical geometric priors at both the latent representation and image reconstruction stages, in a low-overhead manner. Specifically, at the latent representation stage, we adopt a pair of pixel-wise and semantic-wise sphere-to-planar distortion maps to perform affine transformations on the latent representation, thereby incorporating it with spherical properties. Moreover, during the image reconstruction stage, we introduce a geodesic-based resampling strategy, aligning the implicit image function with spherical geometrics without introducing additional parameters. As a result, the proposed FAOR outperforms the state-of-the-art ODI-SR models with a much faster inference speed. Extensive experimental results and ablation studies have demonstrated the effectiveness of our design.

Problem

• Addresses Omni-Directional Image Super-Resolution
• Incorporates spherical geometric priors efficiently
• Achieves faster inference speed in ODI-SR

Innovation

• Implicit image function adaptation
• Spherical geometric priors integration
• Geodesic-based resampling strategy