Three-Dimensional Reconstruction Using Enhanced Shape from Stereo Technique

Three-Dimensional reconstruction is the art of getting a 3D model using flat 2D images, this process becomes too difficult when we use a few number of images like the case of stereo images. In stereo images, pair of images are used to reconstruct the view. The reconstruction process depends on the matching process, in which every feature in the first image is matched with its exact feature in the second image. Matching the features only will affect the smoothness and the density of the reconstructed model, in addition matching the whole features in the first image with the whole features in the second image is an exhausting process, since it wastes the processing time and affects the accuracy of the matching process. In this paper we proposed a novel technique to get a more dense 3D model depending on more accurate matching process, we made the matching process in two levels, the first level is matching the features the second level is matching the pixels , then we used the adaptive window technique to match in a small area rather than matching the whole image. The results showed an essential enhancement in the accuracy of the matching process giving a more dense 3D model.