3D~1V!®DEL RECONSTRUCTION OF COMPLEX OBJECTS FROM MULTIPLE IMAGES

Abstract

This thesis is primarily concerned with the design and construction of image-based object reconstruction system from multiple silhouette images taken by a. natural camera. We briefly explore two techniques of object reconstruction, voxelization and reconstruction based on SOR properties. We propose a view dependent image-based method to recover a 3D-model of generic complex object. We discuss the problem of reconstruction to compute geometric information by finding corresponding feature points among multiple reference images. We exploit the geometry of single axis motion along with the surface of revolution to acquire the surface properties of object from the sequence of multiple homologous images. We further extend our technique to comply with non-homologous images. The constrained and controlled environment will enable us to avoid the need for actual camera calibration