Cálculo de las coordenadas espaciales de una escena 3-D por medio del análisis de imágenes estereoscópicas

Abstract

This paper addresses the reconstruction of a 3-D scene from stereoscopic images containing straight line segments. First the edges of the images are detected based on both Canny´s approach and a heuristic method to track the edge points. Each line is extracted by fitting the pixels to a linear model and measuring the variance which gives an estimate of how well the data adjust to the model. Due to fragmentation of segments a collinearity detection stage is implemented and the corresponding pairs are searched. A Graph theory method is applied to find the right matches. Each node of the graph represents a corresponding pair and the adjacency matrix is built establishing the compatibility between nodes. Then the maximum clique is found and 3-D line coordinates are calculated by means of triangulation. Finally an interpolation stage takes place and a dense depth map is recovered.