Abstract:Omnidirectional images (ODIs) have become increasingly popular, as their large field-of-view (FoV) can offer viewers the chance to freely choose the view directions in immersive environments such as virtual reality. The Möbius transformation is typically employed to further provide the opportunity for movement and zoom on ODIs, but applying it to the image level often results in blurry effect and aliasing problem. In this paper, we propose a novel deep learning-based approach, called \textbf{OmniZoomer}, to incorporate the Möbius transformation into the network for movement and zoom on ODIs. By learning various transformed feature maps under different conditions, the network is enhanced to handle the increasing edge curvatures, which alleviates the blurry effect. Moreover, to address the aliasing problem, we propose two key components. Firstly, to compensate for the lack of pixels for describing curves, we enhance the feature maps in the high-resolution (HR) space and calculate the transformed index map with a spatial index generation module. Secondly, considering that ODIs are inherently represented in the spherical space, we propose a spherical resampling module that combines the index map and HR feature maps to transform the feature maps for better spherical correlation. The transformed feature maps are decoded to output a zoomed ODI. Experiments show that our method can produce HR and high-quality ODIs with the flexibility to move and zoom in to the object of interest. Project page is available at <a class="link-external link-http" href="http://vlislab22.github.io/OmniZoomer/" rel="external noopener nofollow">this http URL</a>.

Smooth, Efficient, and Interruptible Zooming and Panning

Automated Video Looping with Progressive Dynamism

Web-based virtual walkthrough of panoramas

Fast Edit Propagation for 360 Degree Panoramas Using Function Interpolation

High-quality Texture-based Flow Visualization on Surfaces.

Illustrative Motion Smoothing for Attention Guidance in Dynamic Visualizations

Spot-tracking Lens: A Zoomable User Interface for Animated Bubble Charts.

Interactive Localized Liquid Motion Editing

Fast Generation Of Dynamic And Multi-Resolution 360 Degrees Panorama From Video Sequences

Dynamically Tunable Hemispherical Electronic Eye Camera System with Adjustable Zoom Capability

Zooming through History: Active Perceptual Experience Mediated by Technology

Rendering by Manifold Hopping

Pull and Push: Proximity-Aware User Interface for Navigating in 3D Space Using a Handheld Camera

ShellCam: Interactive geometry-aware virtual camera control

Active Animations of Reduced Deformable Models with Environment Interactions

Path‐space Motion Estimation and Decomposition for Robust Animation Filtering

Focus + Context Visualization with Animation

The Signals and Systems Approach to Animation

Fast Construction of Dynamic and Multi-Resolution 360° Panoramas from Video Sequences

Creating dynamic panorama using particle swarm optimization

OmniZoomer: Learning to Move and Zoom in on Sphere at High-Resolution