Matching of omnidirectional images based on the geodesic distance
Ibrahim Guelzim, Amina Amkoui, Lahcen Koutti · 2016
In the omnidirectional vision, the catadioptric sensors causes a non-uniform resolution and geometric distortions in resulting images, hence, the conventional processing methods are not convenient. The aim of this work is to adapt matching method based on the proximity criterion to omnidirectional stereovision. The adapted method based on geodesic distance since the treatments are done on a spherical space. The similarity measures used are mutual information and correlation coefficient. We chose the adapted processing methods rather than conventional ones. We work in the case which the sensor has a slight displacement between two images. The tests are done on normal and noisy images. The results are satisfactory.