Tilt Correction of Panoramic Images for a Holistic Visual Homing Method with Planar-Motion Assumption

Holistic local visual homing based on warping of panoramic images relies some simplifying assumptions about the and environment to make problem more tractable. One these is that are captured flat ground without tilt. While this might be true in environments, it poses a for wider real-world application warping. An extension framework proposed where tilt-corrected used as inputs. The method combines tilt correction with systematic search through hypothetical parameters, using an image distance measure produced by optimization criterion. This not only improves performance tilted images, but also allows good estimation requiring additional sensors or external alignment. Experiments two newly collected databases confirm improved viability tilt-estimation scheme. Approximations tilt-correction transformations multiple direct strategies evaluated respect their runtime vs. quality trade-offs find variant methods which best fulfills requirements practical applications.