High quality pictorial detail acquisition and patching on 3D objects

Sommario non disponibile.

In this paper we propose a new approach for the creation of highly realistic models of real free-form 3D objects. We propose the integration of different tech-niques (automatic 3D scanning, image processing, inverse texture-mapping and surface-based 3D graphics) into a coherent modeling system. The system starts from a 3D mesh (acquired with any of the available 3D scanning technologies) and improves this model with pictorial detail and, if required, high frequency shape de-tail. This added information is obtained by means of a small set of photographs of the object; these images are processed in order to: compute camera calibration and position, remove illumination effects to compute both diffuse reflectance proper-ties and a high resolution surface normal field, and finally to attach the acquired detail on the triangle mesh via standard texture mapping. In the very last phase, we integrate the images into a single texture map by blending the different images, and also detect mesh areas that are not covered by any image. In particular, a new approach is proposed to produce a smooth join between different images that maps on adjacent sections of the surface. This is done by applying an accurate piecewise local registration of the original images and by blending textures. For each mesh face which is on the adjacency border between different observed images, texture blending is operated by resampling a new texture patch obtained as a weighted blend of the corresponding adjacent images. Examples of the results obtained with sample Cultural Heritage objects are pre-sented and discussed.