Color features comparison for segments matching

Christophe Montagne (1) , Sylvie Lelandais (1) , André Smolarz (2)

1 IBISC - Informatique, Biologie Intégrative et Systèmes Complexes
2 M2S - Modélisation et Sûreté des Systèmes

In the field of the localization of in-door mobile robot, a method consists in extracting line segments from two stereoscopic images and in pairing line segments from the first image with those from the second image. With pairs of line segments, it is possible to measure distances between robot and objects of the environment. The pairing phase requires parameters characterizing the line segments which are discriminating, otherwise the line segments can be badly paired. This paper proposes a method of pairing based on two parameters using information color. More precisely, the two parameters use levels from a color axis which is given according to the color contents of the pair of images. The first parameter is a set of homogeneous colors which are in the areas close to a line segment. These colors are obtained from neighboring gray level dependence matrices but color axis levels replace gray levels. The second parameter is a distance between pixels composing the areas close to a line segment. This distance is the earth mover's distance. Results show a good success rate of the two parameters. The homogeneity colors are also more discriminating than the EMD. These results are extended to experimentation on more two hundred pair of images.

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement des images [eess.IV]; Imagerie médicale; Traitement du signal et de l'image [eess.SP]

Christophe Montagne, Sylvie Lelandais, André Smolarz. Color features comparison for segments matching. 3rd European Conference on Color in Graphics, Imaging and Vision (CGIV 2006), Jun 2006, Leeds, United Kingdom. pp.(Elect. Proc.). ⟨hal-00342242⟩