Automatic mapping of uncalibrated pictures on dense 3D point clouds

This paper presents a solution for the automated texturing of models obtained from 3D scanners by using un-calibrated pictures taken with a digital camera. The proposed method exploits calibrated light intensity images given by the scanning device to automatically extract a set of correspondences with the uncalibrated pictures, such correspondences are used to give an accurate estimate for the camera pose. In a second phase the set of calibrated pictures is projected onto the model to build the texture, with a proper merging of overlapping images. The image blending is carried out by weighting the pixels of each picture depending on their depth and normal orientation. Results are proposed for sculpture datasets where a method to assess the accuracy of the mapping is applied.