Skin Detection in frontal-view faces (bibtex)
by Robin Melan
Abstract:
Skin detection plays an important role in a wide range of image processing applications such as image classification, face detection, face tracking, content-based image retrieval, gesture analysis and various human computer interaction domains. In recent years, the number of skin segmentation approaches has grown. However, skin detection remains an open problem due to its challenges illumination, complex background, camera characteristics and ethnicity. This report presents a new model-based approach using classification learners with supervised learning on frontal-view face images. The proposed solution is based on independent pixel classifiers, namely weighted kNN and decision trees. Both classifiers are trained from automatically labeled data and extend it by using Viola-Jones eyes and nose detectors and Active Contour Model (ACM) to extract sample pixels of both skin and non-skin classes. Our evaluation gives a comparative study with baseline state-of-the-art explicit thresholding methods. This methodological approach is seen as a preprocessing step of the following master thesis Automatic human-head and shoulder segmentation of frontal-view face images.
Reference:
Skin Detection in frontal-view faces (Robin Melan), (Walter G. Kropatsch, Nicole M. Artner, eds.), Technical report, PRIP, TU Wien, 2018.
Bibtex Entry:
@TechReport{TR142,
  author =	 "Robin Melan",
  editor = "Walter G. Kropatsch and Nicole M. Artner",
  title =	 "Skin Detection in frontal-view faces",
  institution =	 "PRIP, TU Wien",
  number =	 "PRIP-TR-142",
  year =	 "2018",
  url =		 "ftp://ftp.prip.tuwien.ac.at/pub/publications/trs/tr142.pdf",
  abstract =	 "Skin detection plays an important role in a wide range of image processing applications such
as image classification, face detection, face tracking, content-based image retrieval, gesture
analysis and various human computer interaction domains. In recent years, the number of
skin segmentation approaches has grown. However, skin detection remains an open problem
due to its challenges illumination, complex background, camera characteristics and ethnicity.
This report presents a new model-based approach using classification learners with supervised learning on frontal-view face images. The proposed solution is based on independent
pixel classifiers, namely weighted kNN and decision trees. Both classifiers are trained from automatically labeled data and extend it by using Viola-Jones eyes and nose detectors and Active
Contour Model (ACM) to extract sample pixels of both skin and non-skin classes. Our evaluation gives a comparative study with baseline state-of-the-art explicit thresholding methods.
This methodological approach is seen as a preprocessing step of the following master thesis
Automatic human-head and shoulder segmentation of frontal-view face images.",
}
Powered by bibtexbrowser