Appearance-based face recognition using aggregated 2d gabor features

King Hong Cheung, Jia You, James Liu, W. H. Tony

Research output: Journal article publicationJournal articleAcademic researchpeer-review


Face recognition performed in a controlled environment can be transformed to classical image retrieval/pattern recognition of frontal 2D images of a person, i.e. mug shot. Current holistic appearance based face recognition methods require a high dimensional feature space to attain fruitful performance. We, therefore, propose a relatively low feature dimension scheme to cope with the transformed face recognition problem. Aggregated Gabor filter responses is employed to represent face images. We have conducted experiments on two testing sets of a face image database. Each set contains over 3,000 images of the same 120 subjects and they differ from each other in the preprocessing of image size. We have compared the performance of using our method and PCA and the performance of using L1 and L2-norm as (dis)similarity measures to guide the retrieval process.
Original languageEnglish
Pages (from-to)572-579
Number of pages8
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publication statusPublished - 1 Dec 2004

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)


Dive into the research topics of 'Appearance-based face recognition using aggregated 2d gabor features'. Together they form a unique fingerprint.

Cite this