Motion blur correction on images taken at night from a digitized terrestrial vehicle

This current work investigates the possibility of acquiring images with long exposure time, by defining a method of motion blur correction from photographs taken with the specific imaging equipment of a mobile mapping system. In the proposed approach, we take into consideration both inertial data provided by accelerometers and gyroscopes and spatial data coming from Lidar cloud or 3D models and bringing information about the variations of the scene depth. Our algorithm exploits all useful data provided by the mobile mapping system to compute in the most accurate way the point spread function for each pixel. We propose also a first attempt of blur removing using our non-uniform and spatially-varying blur kernel in each pixel and spatial considerations. Our method is currently tested on blurred and non-noisy images generated by synthetic images computing the real vehicle movement. We eventually discuss how to get a whole image corrected from motion blur and how to improve yet our first results.