Publication detail

Effective Facial Feature Keypoint Detection Using Active Shape Models Composition

PŘINOSIL, J.

Original Title

Effective Facial Feature Keypoint Detection Using Active Shape Models Composition

Czech Title

Efektivní detekce klíčových bodů částí obličeje za použití kompozice aktivních modelů tvaru

English Title

Effective Facial Feature Keypoint Detection Using Active Shape Models Composition

Type

conference paper

Language

en

Original Abstract

This paper deals with the proposal of effective system for facial feature keypoints detection in static images, where a composition of several Active Shape Models ASM is used. Shape of each facial feature (eyes, mouth, etc.) is represented by a minor model of facial feature keypoints and central positions of these minor models in a particular face are represented by a general model. This approach leads to fast convergence to appropriate result and allows using more complex keypoint descriptors extraction method for more reliable detection. In the paper an input image response on a set of the Gabor filters with the GentleBoost classification is used for that. Comparison of proposed system with standard ASM technique represented by stasm library is included in the end of the paper.

Czech abstract

Článek se zabývá návrhem efektivního sytému detekce klíčových bodů jednotlivých částí obličeje ve statickém obraze, přičemž je využito kompozice několika aktivních modelů tvaru ASM. Tvar každé části obličeje (oči, nos, atd.) je reprezentován minoritním modelem klíčových bodů dané části a pozice středů těchto modelů pro příslušný obličej je zastoupena obecným modelem. Tento přístup vede k rychlé konvergenci k požadovaným výsledkům a dovoluje tak použití složitějších metod extrakce deskriptorů příslušných klíčových bodů pro spolehlivější detekci. V tomto článku je použita impulsní odezva vstupního obrazu na sadu Gaborových filtrů společně s klasifikátorem GentleBoost. Na závěr článku je uvedeno srovnání navrženého systému se standardní metodou ASM zastoupenou knihovnou stasm.

English abstract

This paper deals with the proposal of effective system for facial feature keypoints detection in static images, where a composition of several Active Shape Models ASM is used. Shape of each facial feature (eyes, mouth, etc.) is represented by a minor model of facial feature keypoints and central positions of these minor models in a particular face are represented by a general model. This approach leads to fast convergence to appropriate result and allows using more complex keypoint descriptors extraction method for more reliable detection. In the paper an input image response on a set of the Gabor filters with the GentleBoost classification is used for that. Comparison of proposed system with standard ASM technique represented by stasm library is included in the end of the paper.

Keywords

Active Shape Model, facial keypoints, Gabor filters, GentleBoost classifier

RIV year

2010

Released

19.08.2010

Publisher

Asszisztencia Szervezo Kft.

ISBN

978-963-88981-0-4

Book

In Proceeding of the 33rd International Conference on Telecommunications and Signal Processing - TSP 2010

Pages from

200

Pages to

203

Pages count

4

BibTex


@inproceedings{BUT34910,
  author="Jiří {Přinosil}",
  title="Effective Facial Feature Keypoint Detection Using Active Shape Models Composition",
  annote="This paper deals with the proposal of effective system for facial feature keypoints detection in static images, where a composition of several Active Shape Models ASM is used. Shape of each facial feature (eyes, mouth, etc.) is represented by a minor model of facial feature keypoints and central positions of these minor models in a particular face are represented by a general model. This approach leads to fast convergence to appropriate result and allows using more complex keypoint descriptors extraction method for more reliable detection. In the paper an input image response on a set of the Gabor filters with the GentleBoost classification is used for that. Comparison of proposed system with standard ASM technique represented by stasm library is included in the end of the paper.",
  address="Asszisztencia Szervezo Kft.",
  booktitle="In Proceeding of the 33rd International Conference on Telecommunications and Signal Processing - TSP 2010",
  chapter="34910",
  howpublished="electronic, physical medium",
  institution="Asszisztencia Szervezo Kft.",
  year="2010",
  month="august",
  pages="200--203",
  publisher="Asszisztencia Szervezo Kft.",
  type="conference paper"
}