Skip to Main content Skip to Navigation

Visuo-­perceptual validation methods for physically based image synthesis

Abstract : The simulation of physico-realistic materials is a process that requires a lot of computation. Since the images are meant to be seen by human observers, we can use the limitations of their visual system to simplify the rendering model, avoiding redundant information that will not be seen. This is known as perceptual realism. Focusing on the simulation of automobile paint coatings, with special attention to metallic-flaked coatings with a sparkling appearance, we try to improve perceptual realism in two ways: using stereoscopic visualization, to provide additional depth information from binocular disparity; and preserving as much of the original perceptual luminance information as possible. The unlimited luminance levels, or dynamic range, of a real scene must be reduced as an image is processed by lower-dynamic range media throughout the acquisition and visualization chain. To ensure perceptual accuracy throughout this process, we propose a methodology consisting on device characterization, radiometric acquisition, and visuo-perceptual validations. Replacing the human eye by a DSLR camera, as a trichromatic color integrator of radiometric information, we perform visual comparisons of real samples and photographs to estimate the image exposure that maximizes perceptual accuracy under a controlled observation environment. These results are then contrasted with different tone reproduction methods, in order to analyze the effects on texture perception of specific image attributes like exposure, dynamic range, brightness, and contrast. We also propose a full methodology to produce simulations of a real scene, which are radiometrically and colorimetrically comparable to photographs of the same scene. By ensuring that the simulation produces correct images, this methodology lays the foundations for a future integration of our observations into the rendering engine.
Document type :
Complete list of metadata

Cited literature [120 references]  Display  Hide  Download
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, September 12, 2017 - 10:25:15 AM
Last modification on : Wednesday, November 17, 2021 - 12:31:04 PM
Long-term archiving on: : Wednesday, December 13, 2017 - 5:58:28 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01585882, version 1


Victor Medina. Visuo-­perceptual validation methods for physically based image synthesis. Graphics [cs.GR]. Université Paris sciences et lettres, 2016. English. ⟨NNT : 2016PSLEM029⟩. ⟨tel-01585882⟩



Record views


Files downloads