Feng, Jufu
Professor
Research Interests: Pattern recognition, biometric authentication
Office Phone: 86-10-6275 8814
Email: fjf@cis.pku.edu.cn
Feng, Jufu is a professor in the Department of Machine Intelligence, School of EECS. He obtained his B.Sc. in 1989 and Ph.D. in 1997 both from Peking University. His research interests include pattern recognition, machine learning and biometric authentication.
Dr. Feng has published more than 100 research papers in important conferences and journals. He was awarded the 2nd Award for Progress of Science and Technology from Ministry of Education (2000), Pattern Recognition Letter Top Cited Article 2005-2010(2011), and the 2nd Award for Progress of Science and Technology from Ministry of Public Security (2012).
Dr. Feng has more than ten research projects including NSFC, 973 programs, etc. His research achievements are summarized as follows:
1) Fingerprint and Palmprint Recognition: He and his collaborators proposed a minutia extraction method based on Gabor Amplitude-Phase model. The fingerprint/palmprint image is convolved by a complex Gabor filter and then transformed into the phase field and amplitude field. Differing from most existing methods, a minutiae extractor extracts minutiae directly from the Gabor phase field without binarization and thinning and the Gabor amplitude field can be used to measure the credibility of minutiae. He and his collaborators also proposed a series of fingerprint/palmprint matching algorithms. Some proposed algorithms are integrated into the Peking University Automated Fingerprint Identification System (PU-AFIS), which has been widely used in China.
2) Image distance: The distance metric plays a fundamental role in pattern recognition and machine learning. He and his collaborators proposed IMage Euclidean Distance (IMED) which takes into account the spatial relationships of pixels and showed that IMED is equivalent to a translation-invariant transform. Furthermore, they proved any translation invariant metric is equivalent to a closed-form translation invariant transform. They also proposed a fast implementation of IMED and a transform domain metric learning algorithm.
