Research On Finger-knuckle-print Recognition Based On Steerable Filter

Biometrics uses human behavior or physical characteristics for personal identifica-tion, which is based on digital image processing and uses the method from pattern recog-nition. With the high recognition accuracy and robust identification property compared totraditional methods, it has received much attention in recent years. A number of biomet-ric traits such as finger print, human face and iris has been exhaustively investigated byresearchers and has also been successfully applied into industry.Among various kinds of biometric identifiers, hand-based biometrics has been at-tracting considerable attention over recent years for their property of easy-access, robust-ness and high user acceptance. Recently, it has been found that the finger-knuckle-print(FKP), which refers to the inherent skin patterns of the outer surface around the pha-langeal joint of one’s finger, has high capability to discriminate diferent individuals. Atthe same time, FKP is hard to be abraded and people rarely leave it remains on the stufsurface, which makes it an emerging biometric identifier.A typical FKP recognition system consists of four modules: image acquisition, pre-processing, feature extraction and feature matching. Progress of FKP recognition is firstrevisited in this paper, and state-of-art methods used in feature extraction and matchingis then reviewed in details. The traditional Competitive Coding method which is widelyused in palm recognition is first applied in FKP recognition, then several improvementsbased on the own characteristics of FKP is made on it. Log-Gabor filter is used forimage filtering instead of the Gabor filter to remove afect the DC, the number of filtersin the filter bank for competitive is also reduce to four, which could speed up the featureextraction and matching process with little afect to recognition accuracy.For further analysis of local orientation feature of FKP and improvements of recog-nition accuracy, a novel FKP recognition method using the Adaptive Steerable OrientationCoding (ASOC) is proposed. High order steerable filters are first employed to extract thecontinuous orientation feature map, then we use multilevel histogram threshold methodto quantize the feature map adaptively and the discrete orientations are used for codinga FKP image. Furthermore, we measure the similarity between two coded FKP imagesby designing an efective angular matching function. Experimental results on the PolyUFKP database demonstrate that ASOC can extract the local orientation of FKP robustly and accurately, and4orientation classes are well enough to represent the FKP line featurefor recognition. Compared with the state-of-art coding based methods, ASOC achievesbetter performance on the PloyU FKP database.How the type of FKP afects the recognition accuracy is investigated at the end of thepaper, further fusion on diferent type of finger from the same individual is also discussed.