• Refine Query
  • Source
  • Publication year
  • to
  • Language
  • 1
  • Tagged with
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • About
  • The Global ETD Search service is a free service for researchers to find electronic theses and dissertations. This service is provided by the Networked Digital Library of Theses and Dissertations.
    Our metadata is collected from universities around the world. If you manage a university/consortium/country archive and want to be added, details can be found on the NDLTD website.
1

Face recognition under transformations of intensity

Zhang, Yan 12 April 2018 (has links)
Variable illumination intensity on unsegmented human face is one of the most challenging problems for reliable face recognition. When the illumination conditions are uncontrolled in a scene, the output of face images will become quite different. We applied a new locally adaptive contrast-invariant filter (LACIF) method in face recognition. Multiplicative and additive transformations of intensity over face images are combined by this filter. And three correlations are calculated by a nonlinear way. The correlation peaks show that LACIF is invariant under a uniform intensity transformation over the face. An extended method based on LACIF is also applied in face recognition. In this method, a linear intensity gradient across the face is considered. A set of basis face images is established. And five correlation planes are combined in a nonlinear way. Thousands of computer simulations are performed to test the face recognition capability. Results show that the discrimination is excellent. We also applied traditional and extended LACIF methods in face recognition with real-world environment. Results with actual experiments demonstrate these methods are effective and robust in real-world face objects. Face recognition is invariant under the intensity transformation. The discrimination capability is good. / Nous avons appliqué une nouvelle méthode de LACIF (locally adaptive contrast-invariant filter ) dans la reconnaissance de visage. Des transformations multiplicatives et additives des images finies de visage d'intensité sont combinées par ce filtre. Les crêtes de corrélation prouvent que LACIF est invariable sous une transformation uniforme d'intensité au-dessus du visage. Une méthode prolongée basée sur LACIF est également appliquée dans la reconnaissance de visage. Dans cette méthode, un gradient linéaire d'intensité à travers le visage est considéré. Un ensemble d'images base de visage est établi. Des milliers de simulations sur ordinateur sont effectués pour examiner les possibilités la reconnaissance de visage. Les résultats prouvent que la discrimination est excellente. Nous avons également appliqué des méthodes traditionnelles et prolongées de LACIF dans la reconnaissance de visage avec l'environnement réel. Les résultats avec des expériences réelles démontrent la reconnaissance de visage est invariable sous la transformation d'intensité.

Page generated in 0.0415 seconds