Aplicación de una metodología de evaluación de sistemas de emparejamiento en visión tridimensional

Abstract

Application of a methodology to evaluate matching methods in three-dimensional vision. In this article a methodology to compare and evaluate matching methods in stereopsis is partially applied. Also, the following implementations of matching methods based on the human visual system have been developed: the Marr and Poggio's model (1979), the Pollard, Mayhew and Frisby's PMF algorithm (1985), the Kim and Aggarwal's algorithm (1987) and the Rubio de Lemus' RP algorithm (1991). With the application of the methodology, its possibilities to produce a series of results about the algorithms have been confirmed due to the fact that it evaluates their most relevant aspects. To this aim, these methods have been compared through variables at the algorithm level and the implementation level. The results have been statistically analysed, concluding that RP and PMF are more efficient than the other stereo algorithms.