Non-professionals may someday be able to create high-quality video panoramas using multiple cameras with the help of an algorithm developed by a team of Disney researchers.

Their method smooths out the blurring, ghosting and other distortions that routinely occur when video feeds from unstructured camera arrays are combined to create a single panoramic video. The algorithm corrects for the apparent difference in position of an object caused by parallax (different camera angles) and image warping that occurs because of slight timing differences between cameras, both of which lead to visible discontinuities, ghosting, and other imperfections in existing approaches.

Linking random smartphones

The researchers have demonstrated their technique using as many as 14 various types of cameras, generating panoramic video with tens of megapixels to more than 100 megapixels.

“We can foresee a day when just about anyone could create a high-quality video panorama by setting up a few video cameras or even linking several smartphones, just as many people today can easily create a still photo panorama with their smartphones,” said Alexander Sorkine-Hornung, a senior research scientist at Disney Research Zürich, who collaborated with colleagues at ETH Zürich and Disney Imagineering on the study.

Their open-access findings are being presented at EUROGRAPHICS 2015, the Annual Conference of the European Association for Computer Graphics, May 4–8, in Zürich, Switzerland.

Though some professional methods using pre-calibrated camera arrays exist for creating video panoramas, the Disney team focused on combining videos from multiple cameras that have overlapping visual fields, but are not precisely positioned and are not perfectly synchronized.

Algorithms that correct for position, alignment, and timing

Their technique automatically analyzes the images from the cameras to estimate the position and alignment of each camera, which eliminates the need to use special or manual calibration techniques, and allows for a flexible positioning of the cameras.

The algorithm corrects for differences in parallax that create ghosting and other disturbing effects in the areas of the panorama where images from separate cameras are stitched together. It also detects and corrects for image warping — wavy lane markings on roads, or buildings that appear to bend over — that occurs when images are stitched together. The technique also compensates for slight differences in the timing of frames between cameras, which otherwise causes jitter and other artifacts in the image.

Disney Research Hub | Panoramic Video from Unstructured Camera Arrays

---

Abstract of Panoramic Video from Unstructured Camera Arrays

We describe an algorithm for generating panoramic video from unstructured camera arrays. Artifact-free panorama stitching is impeded by parallax between input views. Common strategies such as multi-level blending or minimum energy seams produce seamless results on quasi-static input. However, on video input these approaches introduce noticeable visual artifacts due to lack of global temporal and spatial coherence. In this paper we extend the basic concept of local warping for parallax removal. Firstly, we introduce an error measure with increased sensitivity to stitching artifacts in regions with pronounced structure. Using this measure, our method efficiently finds an optimal ordering of pair-wise warps for robust stitching with minimal parallax artifacts. Weighted extrapolation of warps in non-overlap regions ensures temporal stability, while at the same time avoiding visual discontinuities around transitions between views. Remaining global deformation introduced by the warps is spread over the entire panorama domain using constrained relaxation, while staying as close as possible to the original input views. In combination, these contributions form the first system for spatiotemporally stable panoramic video stitching from unstructured camera array input.