While both work with images, computer graphics and computer vision are inverse problems. Computer graphics starts traditionally with input geometric models and produces image sequences. Computer vision starts with input image sequences and produces geometric models. In the last few years, there has been a convergence of research to bridge the gap between the two fields. This convergence has produced a new field called Image-based Rendering and Modeling (IBMR). IBMR represents the effort of using the geometric information recovered from real images to generate new images with the hope that the synthesized ones appear photorealistic, as well as reducing the time spent on model creation. In this dissertation, the capturing, geometric and photometric aspects of an IBMR system are studied. A versatile framework was developed that enables the reconstruction of scenes from images acquired with a handheld digital camera. The proposed system targets applications in areas such as Computer Gaming and Virtual Reality, from a lowcost perspective. In the spirit of IBMR, the human operator is allowed to provide the high-level information, while underlying algorithms are used to perform low-level computational work. Conforming to the latest architecture trends, we propose a streaming voxel carving method, allowing a fast GPU-based processing on commodity hardware.