3D Reconstruction Using Level-Based Approach in Stereopsis

Abstract

We present preliminary results of our work on an alternative approach to address correspondence in stereopsis. Accurate correspondence between two images is required to determine depth and hence, the 3D model, out of two 2D images. To achieve this, we studied the relationship among disparity, depth, camera features and data gathered from output images. We propose a level-based approach to enhance the correspondence accuracy. The results show that the algorithm gives accurate results compared to the measurements in the camera set-up