Colour in Context
Research group
Computer Vision Center

Ordinal pairwise method for natural images comparison

Perception, European Conference on Visual Perception (ECVP´09), Volume 38, page 180 - 2009
IF: 1.462. area: PSYCHOLOGY. Quartile: 3.
Download the publication : VPV09.pdf [9Ko]  
We developed a new psychophysical method to compare different colour appearance models when applied to natural scenes. The method was as follows: two images (processed by different algorithms) were displayed on a CRT monitor and observers were asked to select the most natural of them. The original images were gathered by means of a calibrated trichromatic digital camera and presented one on top of the other on a calibrated screen. The selection was made by pressing on a 6-button IR box, which allowed observers to consider not only the most natural but to rate their selection. The rating system allowed observers to register how much more natural was their chosen image (eg, much more, definitely more, slightly more), which gave us valuable extra information on the selection process. The results were analysed considering both the selection as a binary choice (using Thurstone´s law of comparative judgement) and using Bradley-Terry method for ordinal comparison. Our results show a significant difference in the rating scales obtained. Although this method has been used in colour constancy algorithm comparisons, its uses are much wider, eg to compare algorithms of image compression, rendering, recolouring, etc.

Images and movies

 

BibTex references

@InProceedings\{VPV2009,
  author       = "Javier Vazquez-Corral and C. Alejandro Parraga and Maria Vanrell",
  title        = "Ordinal pairwise method for natural images comparison",
  booktitle    = "Perception, European Conference on Visual Perception (ECVP\´09)",
  volume       = "38",
  pages        = "180",
  year         = "2009",
  abstract     = "We developed a new psychophysical method to compare different colour appearance models when applied to natural scenes. The method was as follows: two images (processed by different algorithms) were displayed on a CRT monitor and observers were asked to select the most natural of them. The original images were gathered by means of a calibrated trichromatic digital camera and presented one on top of the other on a calibrated screen. The selection was made by pressing on a 6-button IR box, which allowed observers to consider not only the most natural but to rate their selection. The rating system allowed observers to register how much more natural was their chosen image (eg, much more, definitely more, slightly more), which gave us valuable extra information on the selection process. The results were analysed considering both the selection as a binary choice (using Thurstone\´s law of comparative judgement) and using Bradley-Terry method for ordinal comparison. Our results show a significant difference in the rating scales obtained. Although this method has been used in colour constancy algorithm comparisons, its uses are much wider, eg to compare algorithms of image compression, rendering, recolouring, etc.",
  ifactor      = "1.462",
  quartile     = "3",
  area         = "PSYCHOLOGY",
  url          = "http://cat.cvc.uab.es/Public/Publications/2009/VPV2009"
}

Other publications in the database

 © 2008 Colour in context Group | Computer Vision Center. All rights reserved | Contact webmaster |  Last updated: Monday 11 May 2009     eXTReMe Tracker