360-degree panoramic video generation using virtual and actual multiple cameras

This paper proposes a technique for 360-degree panoramic video generating with multiple actual and virtual cameras in real-time. Our method acquires images consisting of real from physical by GPU-based texture encoding. Then, our generates ORB feature detection matching to images, automatic homography estimation based on the random absent consensus algorithm, multi-band blending produces high-resolution stream Nvidia encoder Hardware accelerated From experimental results, we verified that takes at least 45fps 2K resolution 33fps 4K when transmitting head mounted display stereo, enabling real-time generation as well transmission.