Assessing the Metrological Performance of DIC Applied on RGB Images - Université Clermont Auvergne Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

Assessing the Metrological Performance of DIC Applied on RGB Images

Benoît Blaysat
  • Fonction : Auteur
Michel Grédiac
  • Fonction : Auteur
  • PersonId : 1005345

Résumé

Thanks to the rise of digital reflex cameras, RGB CMOS sensors present nowadays interesting features, at relatively low prices. For instance, the Canon 5DSR sensor displays 50 million pixels. Nevertheless, such cameras being dedicated to photography and not to measurement, their high pixel density makes the associated sensor noise to be relatively high compared to the sensor noise of measurement cameras. Moreover, a Bayer filter mosaic is added in front of the pixels of such sensors in order to produce RGB images by interpolation. Embedded camera processors demosaice these data to elaborate full RGB pictures. Additional post-processing steps, including white balance and gamma correction, are also used in order to prevent any color shift and to give a pleasing-to-the-eye output. These post-processing steps considerably transform the raw sensor data to produce the output images. The fundamental assumption of DIC, namely the conservation of brightness for a given material point, may thus be violated. In this context, we propose in this study a consistent use of the separate channels of RGB sensors for the DIC minimization. The metrological performance of the DIC measurement using this approach is discussed and compared to that dealing with “classic” gray level images.
Fichier non déposé

Dates et versions

hal-01500324 , version 1 (03-04-2017)

Identifiants

Citer

Benoît Blaysat, Michel Grédiac, Frédéric Sur. Assessing the Metrological Performance of DIC Applied on RGB Images. iDICS 2016 Conference, Nov 2016, Philadelphia, United States. pp.109-111, ⟨10.1007/978-3-319-51439-0_26⟩. ⟨hal-01500324⟩
285 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More