Machine Recognition in Complex Domain - High Dimensional Neurocomputing: Growth, Appraisal and Applications - page 158

Fig. 7.16 Occluded images which are only recognized by C ICA -based recognition system, but can
not be recognized by R ICA and any PCA -based system
Fig. 7.17 Blurred images which are not recognized with any feature extraction technique
Fig. 7.18 Blurred images which are only recognized by C ICA -based recognition system, but can
not be recognized by R ICA and any PCA -based system
Fig. 7.19 Blurred images which are not recognized by any PCA -based recognition system
observed that only C ICA -based recognition system can correctly classify these faces
while other techniques including R ICA failed to do so. Further Fig. 7.19 presents
the set of images possessing lesser blurring in comparison to Fig. 7.18 are again not
recognized by any PCA -based recognition system but they are well recognized by
R ICA and C ICA -based recognizer. It turns out the fact from these observations that
R ICA is better than both PCA but C ICA is best among all. Thus, C ICA provides
more discriminating power among the classes than any other technique, it is due to
robust class distinctiveness yielded by C ICA -based recognition system.