3D Object Model Acquisition from Silhouettes

This thesis proposes an approach for acquiring a 3D object model from silhouettes. The 3D object model we propose is a set of object’s properties which are necessary for reproducing the object’s appearance. The properties in the 3D object model are categorized into three properties; photometry modeled by reflection properties, geometry modeled by object’s shape, and motion modeled by a sequential object’s pose. The 3D object model has two advantages, portability and suitability for preservation. These advantages push a widespread usage of the 3D object model. We categorize the usage into two types; applications which are putting importance on the portability, and applications which are putting importance on preserving objects’ appearance. We can categorize the model acquisition from two types of view; computation time and observation time for model acquisition. The computation/observation time affect on a lighting environment we can use; whether lighting environment is known or unknown, controllable or uncontrollable. There is a relationship between the computation/observation time and controllability of lighting environment. The main contribution of this thesis is to make the relationship clear through the following three research topics. We first propose a method to acquire the shape and reflection properties under unknown and varying lighting environment. The method has flexibility for parallels processing which enables real-time processing. Previously proposed methods had acquired reflection properties by using an object’s shape. Our method does not use the object’s shape and acquires both the shape and the reflection properties. Our method acquires the object’s shape from silhouettes, using the volume intersection method. The shape is acquired as a set of voxels and named a visual hull. Our method also acquires the reflection properties of each voxel. Previously proposed methods, which acquired both the shape and the