# Moving in a 360 World: Synthesizing Panoramic Parallaxes from a Single Panorama

**Authors:** Ching-Yu Hsu, Cheng Sun, Hwann-Tzong Chen  
**arXiv:** [2106.10859](https://arxiv.org/abs/2106.10859)  
**Date:** 2021-06-21  
**Summary page:** https://www.summarizepaper.com/en/paper/2106.10859v1/

## Key points

- Authors Ching-Yu Hsu, Cheng Sun, and Hwann-Tzong Chen introduce OmniNeRF, a method for parallax-enabled novel panoramic view synthesis
- OmniNeRF allows generation of panorama images from unknown viewpoints using just a single equirectangular image as training data
- Key innovation is the ability to seamlessly project between 3D and 2D panoramic coordinates at different virtual camera positions
- Demonstrated effectiveness through experiments on synthetic and real-world datasets, showcasing high-quality panoramic image generation from single input sources
- Represents a significant advancement in view synthesis field with promising implications for immersive visual experiences and realistic scene reconstructions

## Summary

In their paper "Moving in a 360 World: Synthesizing Panoramic Parallaxes from a Single Panorama," authors Ching-Yu Hsu, Cheng Sun, and Hwann-Tzong Chen introduce OmniNeRF, a groundbreaking method for parallax-enabled novel panoramic view synthesis. This innovative approach allows for the generation of panorama images from unknown viewpoints using just a single equirectangular image as training data. The key innovation of OmniNeRF is its ability to seamlessly project between 3D and 2D panoramic coordinates at different virtual camera positions, resulting in accurate estimations of new viewing angles with convincing renderings that exhibit the coveted parallax effect. The effectiveness of OmniNeRF is demonstrated through experiments on both synthetic and real-world datasets, showcasing its ability to generate high-quality panoramic images from single input sources. This research represents a significant advancement in the field of view synthesis and offers promising implications for applications requiring immersive visual experiences and realistic scene reconstructions.