Panoramic stereo imaging system for efficient mosaicking: parallax analyses and system design

Area of Science:

• Computer Vision
• Virtual Reality
• Optical Systems

Background:

• Mosaicking panoramic stereo images for virtual reality (VR) applications is computationally intensive, even with advanced algorithms.
• Current methods struggle with efficiency when combining raw images from distributed devices with varying perspectives.

Purpose of the Study:

• To enhance the efficiency of image mosaicking for VR by optimizing the image capturing system.
• To propose new metrics for assessing the mosaicking capability of image capture setups.

Main Methods:

• Introduction of two parallax factors: peak parallax and deviation of parallaxes, to quantify mosaicking capability.
• Analysis through controlled variable manipulation and numerical computation to establish relationships between parallax factors and system design parameters.

Main Results:

• Revealed rules governing the interplay between parallax factors and design parameters of the capturing system.
• Simulation-based validation identified key design strategies for improved mosaicking.

Conclusions:

• A general design strategy involving large capturing distances, increased camera numbers, compact arrangements, and moderate overlaps is recommended.
• Optimized capturing systems facilitate efficient mosaicking, enabling potential real-time VR applications.