Biclustering-based iterative segmentation of human face images for facial feature extraction

Debby D. WANG, Haoran XIE, Fu Lee WANG, Ran WANG, Xuefei ZHE, Hong YAN

Research output: Chapter in Book/Report/Conference proceedingChapters

Abstract

With the rapid development of biclustering techniques in machine learning and data mining, such techniques have been successfully applied to practical problems such as gene expression analysis, text mining, collaborative filtering and market analysis. In this work, biclustering techniques were applied to segmentation of gray-scale human face images. A biclustering-based framework (BISA), which iteratively partitions an image into subimages/regions in the SVD subspaces and retains those passing the threshold test as effective regions (ERs), was proposed. After the third iteration of BISA in our experiments, most of important facial feature areas were captured and outputted as ERs, which can be further handled by feature-extraction or contour-detection tools. Overall, the proposed framework is useful and efficient in human face detection and facial feature area extraction, and it welcomes other biclustering methods as components for multi-purpose applications. Copyright © 2016 by IEEE.
Original languageEnglish
Title of host publicationProceedings of the 2016 IEEE Region 10 Conference (TENCON)
Place of PublicationDanvers, MA
PublisherIEEE
Pages1126-1129
ISBN (Print)9781509025961
DOIs
Publication statusPublished - 2016

Citation

Wang, D. D., Xie, H., Wang, F. L., Wang, R., Zhe, X., & Yan, H. (2016). Biclustering-based iterative segmentation of human face images for facial feature extraction. In Proceedings of the 2016 IEEE Region 10 Conference (TENCON) (pp. 1126-1129). Danvers, MA: IEEE.

Keywords

  • Face
  • Erbium
  • Facial features
  • Image segmentation
  • Feature extraction
  • Ear
  • Gene expression

Fingerprint Dive into the research topics of 'Biclustering-based iterative segmentation of human face images for facial feature extraction'. Together they form a unique fingerprint.