Xiaoyun Yuan bio photo

Dr. Xiaoyun Yuan
Associate Professor

Email

Github

Google Scholar

Computational Imaging with Camera Arrays

Camera array-based imaging uses a large number of inexpensive cameras to create virtual cameras that outperform real ones. They can function in many ways by changing the arrangement and aiming of the cameras. We developed an unstructured gigapixel array camera (UnstructuredCam), beyond the resolution of a single camera and human visual perception, which aims to capture the large-scale dynamic scene with both wide-FoV and high-resolution. We build a prototype camera array using one global camera and 19 local cameras. All the cameras are mounted on gimbals for adjusting the camera poses.

Left is the captured videos, and right is the recorded screens. It was recoreded in Nanshan I park, Shenzhen. You can choose place to zoom in to see the details interactively. The blue road sign is about 300 m away, the chinese and english characters are very clear. The license plate of the car in 150 m is clearly distinguishable.


By proposing the unstructured embedding algorithm, our UnstructuredCam is robust to local camera movement, loss, and addition because it could react and recover quickly with online recalibration. During realtime streaming, the local camera can be adjusted and online calibrated very quickly. Here, we move the camera from the red position to the yellow postion.


Our UnstructuredCam can also be extended for distant large-scale 3d videography, enabling an unprecedented VR experience:


Based on the UnstructuredCam We built the PANDA and GigaMVS gigapixel dataset (Click me) using our array camera to promote the study of large-scale, long-range, multi-target visual analysis centered on human behavior. More gigapixel videography captured by our UnstructuredCam: