• DocumentCode
    3658946
  • Title

    Physically based image synthesis of materials : a methodology towards the visual comparison of physical vs. virtual samples

  • Author

    Victor Medina;Dominique Lafon-Pham;Alexis Paljic;Emmanuelle Diaz

  • Author_Institution
    Peugeot-Citroen Automobiles S.A., France
  • fYear
    2015
  • fDate
    8/1/2015 12:00:00 AM
  • Firstpage
    1
  • Lastpage
    6
  • Abstract
    The assessment of images of complex materials on an absolute scale is difficult for a human observer. Comparing physical and virtual samples side-by-side simplifies the task by introducing a reference. The goal of this article is to study the influence of image exposure on the perception of realism on images of paint materials containing sparkling metallic flakes. We use a radiometrically calibrated DSLR camera to acquire high resolution raw photographs of our physical samples which provide us with radiometric information from the samples. This is combined with the data obtained from the calibration of a stereoscopic display and shutter glasses to transform the raw photographs into images that can be shown by the display, controlling the colorimetric output signal. This ensures that we can transform our data back and forth between a radiometric and a colorimetric representation, minimizing the loss of information throughout the chain of acquisition and visualization. In this article we propose a paired comparison scenario that improves the results from our previous work, focusing on three main aspects: stereoscopy, exposure time, and dynamic range. Our results show that observers consider stereoscopy as the most important factor of the three for judging the similarity of these images to the reference, followed by exposure time and dynamic range, which supports our claims from previous research.
  • Keywords
    "Image color analysis","Visualization","Observers","Stereo image processing","Radiometry","Lighting","Paints"
  • Publisher
    ieee
  • Conference_Titel
    Colour and Visual Computing Symposium (CVCS), 2015
  • Type

    conf

  • DOI
    10.1109/CVCS.2015.7274878
  • Filename
    7274878