Partially occluded ear recognition based on local features

A local feature based approach was proposed for ear recognition under partial occlusion.Firstly,the Gabor filter is applied for feature extraction.Because the Gabor feature vector is of high dimension,kernel Fisher discriminant analysis(KFDA) is used for dimension reduction as well as class separability enhancement.Based on investigations on the different discriminating ability of sub-regions in ear images,a sub-region and probability based model is proposed for recognition.Experimental results on the USTB ear image database show that ear recognition based on the features extracted from Gabor filtered images performs better than that based on the features extracted from the original images,and the local features based strategy gets a higher recognition rate than the whole image based strategy for recognition.