Image attachment using fuzzy-genetic algorithms
Barna Reskó, Péter Köröndi, Zoltán Petres, J.F. Bourges, Hideki Hashimoto · 2005
This paper gives contribution to the construction of an image mosaic by pasting together a sequence of photographs taken by a regular digital camera from the same position but in different directions. In order to find the best match between two overlapping images, a fuzzy-genetic based algorithm is presented. A translation vector with a minimal difference between the common picture parts has to be found. The difference is defined using the fuzzyfied contours of each images. The search of the optimal translation is considered as a minimization process. This paper proposes a method that uses genetic algorithms to find the minimum point corresponding to the optimal translation. Using the fuzzy based comparison of the two images yields a fast fitness function, while the implementation of the proposed method in a genetic hardware has a low and variable computational cost. The precision of the method is inverse-proportional with the computational time, which makes it ideal to be used in real time systems.