This work presents an information fusion mechanism for image segmentation using multiple cues. Initially, a fuzzy clustering of each cue space is performed and corresponding membership functions are produced on the image coordinates space. The latter include complementary as well as redundant information. A fuzzy inference mechanism is developed, which exploits these characteristics and fuses the membership functions. The produced aggregate membership functions represent objects, which bear combinations of the properties specified by the cues. The segmented image results after post-processing and defuzzification, which involves majority voting. A fuzzy rule based merging algorithm is finally proposed for reducing possible oversegmentation. Experimental results have been included to illustrate the steps and the efficiency of the algorithm.
Fuzzy-logic based information fusion for image segmentation
IEEE International Conference on Image Processing 2005 ; 2 ; II-1210
2005-01-01
534485 byte
Conference paper
Electronic Resource
English
Fuzzy-logic Based Information Fusion for Image Segmentation
British Library Conference Proceedings | 2005
|A fuzzy-logic based approach to color segmentation
SPIE | 2013
|Fuzzy-logic-based sensor fusion of images
SPIE | 2002
|