Some recent results in camera calibration and shape reconstruction

Опубликовано 6 сентября 2016, 6:06

In this talk I will give an overview of some recent work done by my research group in the area of camera calibration and 3D reconstruction. In terms of calibration, I will both address photometric and geometric calibration issues.  We have recently developed a robust photometric calibration procedure that estimates exposure changes, the camera response curve and vignetting from images.  I will also present a simple linear method for fish-eye and cata-dioptric camera calibration and present a new RANSAC approach suited for dealing with quasi-degenerate data.  In terms of reconstruction, I will first present an 3D shape reconstruction approach that uses both silhouette and photoconsistency constraints.  In our approach silhouette constraints are enforced exactly, while photoconsistency and smoothness are optimized.  Finally, I will address the problem of articulated structure and motion recovery from a single camera stream.