Title :
A method of extracting visual main skeleton based on cognition theory
Author :
Xu, Gang ; Lei, Yu-qing
Author_Institution :
Dept. of Electr. & Electron. Eng., North China Electr. Power Univ., Beijing
Abstract :
The paper proposes a method for extracting visual main skeleton based on cognition theory about salience of visual parts, which integrates the advantages of the visual main partspsila reliability for object recognition and the skeletonpsilas reduced-dimension for object representation. Because it can simplify skeleton structure and curve shape and make the results of extraction and description in accord with human visual perception, the method not only has good noise elimination effect, but also can be good at solving recognition difficulties aroused by fuzzy boundaries introduced in image segmentation. At the same time, since it can reduce data complexity and quantity in image description, the method can greatly improve the speed and accuracy of automatic recognition, which are good for skeleton representation and object reconstruction based on the skeleton. The experimental results demonstrate that the method is valid.
Keywords :
cognition; feature extraction; image reconstruction; image representation; image segmentation; image thinning; object recognition; cognition theory; curve shape; data complexity reduction; fuzzy boundaries; human visual perception; image description; image segmentation; noise elimination; object recognition; object reconstruction; object representation; skeleton structure; visual main skeleton extraction; visual parts; Cognition; Data mining; Humans; Image recognition; Noise shaping; Object recognition; Reliability theory; Shape; Skeleton; Visual perception; Discrete curve evolution; Image and graphics; Salience of visual parts; Visual main skeleton; Visual perception;
Conference_Titel :
Machine Learning and Cybernetics, 2008 International Conference on
Conference_Location :
Kunming
Print_ISBN :
978-1-4244-2095-7
Electronic_ISBN :
978-1-4244-2096-4
DOI :
10.1109/ICMLC.2008.4620866