Attribute prediction from near-infrared ocular images
An iris recognition system typically captures an image of the ocular region in the near infrared (NIR). The use of NIR imaging facilities the extraction of texture even from dark color irides (e.g., brown eyes). Our area of research focuses on the prediction of attributes (i.e, gender, race) from the captured NIR ocular image as well as analyzing the covariate impact (i.e., image blur, image scale) on attribute prediction accuracy. Successful attribute prediction can increase the recognition accuracy of an iris recognition system, glean aggregate demographic information from a biometric dataset, provide a semantic description of the individual (i.e., “middle-aged Asian woman with light colored eyes”), reduce database search space, exclude a suspect from further criminal investigation and predict attributes from poor quality images where traditional matchers are likely to fail.
D. Bobeldyk and A. Ross, “Predicting Soft Biometric Attributes from 30 Pixels: A Case Study in NIR Ocular Images,” Proc. of IEEE Winter Applications of Computer Vision Workshops (WACVW), (Hawaii, USA), January 2019.
D. Bobeldyk and A. Ross, “Analyzing Covariate Influence on Gender and Race Prediction from Near-Infrared Ocular Images,” IEEE Access, Vol. 7, pp. 7905-7919, 2019.
D. Bobeldyk and A. Ross, "Predicting Eye Color from Near Infrared Iris Images," Proc. of 11th IAPR International Conference on Biometrics (ICB 2018), (Gold Coast, Australia), February 2018.
D. Bobeldyk and A. Ross, "Iris or Periocular? Exploring Sex Prediction from Near Infrared Ocular Images," Proc. of the 15th International Conference of the Biometrics Special Interest Group (BIOSIG), (Darmstadt, Germany), September 2016.