Colour in Context
Research group
Computer Vision Center

Perceived contrast of natural images is mediated by higher- order image statistics

Tom Troscianko, M. Chirimuuta, C. Alejandro Parraga, D.J. Tolhurst
Perception, European Conference on Visual Perception (ECVP´00), Volume 29, page 82b - 2000
IF: 1.193. area: PSYCHOLOGY.
Download the publication : p173.pdf [28Ko]  
Previous research (Párraga et al, 2000 Current Biology 10 35 - 38) has indicated that human vision is optimised for processing the information in natural visual images. We defined `natural´ as having a normal value of the slope of the Fourier amplitude spectrum (the 1/f slope). The task used was the discrimination of objects subtly varying in shape. The results suggested (a) that object discrimination was optimal for natural values of slope; and (b) that a simple model, based on contrast discrimination mediated by local spatial-frequency detecting units, successfully predicted object discrimination performance. We now ask how this may relate to the perception of overall image contrast. For a variety of natural images, we varied the spectral slope and found that, in general, maximal contrast (as indicated by a matching task), occurs for natural values of spectral slope. This remains the case even when the natural slope deviates from the common value of around -1.2. However, this peak in perceived contrast for natural slopes is removed by phase-scrambling the image. Since this destroys higher-order statistics, while leaving the Fourier spectrum unchanged, we conclude that perceived contrast is mediated by higher-order statistics of images.

Images and movies

 

BibTex references

@InProceedings\{TCP2000,
  author       = "Tom Troscianko and M. Chirimuuta and C. Alejandro Parraga and D.J. Tolhurst",
  title        = "Perceived contrast of natural images is mediated by higher- order image statistics",
  booktitle    = "Perception, European Conference on Visual Perception (ECVP\´00)",
  volume       = "29",
  pages        = "82b",
  year         = "2000",
  abstract     = "Previous research (P\'arraga et al, 2000 Current Biology 10 35 - 38) has indicated that human vision is optimised for processing the information in natural visual images. We defined `natural\´ as having a normal value of the slope of the Fourier amplitude spectrum (the 1/f slope). The task used was the discrimination of objects subtly varying in shape. The results suggested (a) that object discrimination was optimal for natural values of slope; and (b) that a simple model, based on contrast discrimination mediated by local spatial-frequency detecting units, successfully predicted object discrimination performance.
We now ask how this may relate to the perception of overall image contrast. For a variety of natural images, we varied the spectral slope and found that, in general, maximal contrast (as indicated by a matching task), occurs for natural values of spectral slope. This remains the case even when the natural slope deviates from the common value of around -1.2. However, this peak in perceived contrast for natural slopes is removed by phase-scrambling the image. Since this destroys higher-order statistics, while leaving the Fourier spectrum unchanged, we conclude that perceived contrast is mediated by higher-order statistics of images.",
  ifactor      = "1.193",
  area         = "PSYCHOLOGY",
  url          = "http://cat.uab.cat/Public/Publications/2000/TCP2000"
}

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