Generalized transform of Hough

The Transformée generalized of Hough is a technique of Pattern recognition used for the treatment of digital images. Developed in 1972 per R. Duda and P. Hart, it makes it possible to extend the principle of the Transformée of Hough for arbitrary forms ( i.e. without simple analytical representation).

Principle

The generalized transform of Hough functions on the same principle as the transformed of Hough: one seeks the presence of a curve, characterized by a certain number of parameters; each point of the analyzed image “vote” for the whole of the sets of parameters generating of the curves to which it belongs. The presence of a required curve is characterized by a set of parameters having a high score.

Algorithm

Particular case: orientation and fixed size

General case

Advantages and Disadvantages

advantages: simple concept, speed (by stochastic sampling), robustness with the noise, easily extensible with other fields that the imagery.

disadvantages: problem of the homogeneity of space, its quantification; problem of the proof of convergence of the algorithm

References

http://www.cse.unr.edu/~bebis/CS791E/Notes/GeneralizedHough.pdf#search=%22generalized%20hough%20transform%22

Random links: