Color Constancy for Three-Dimensional Objects

Supplemental Material

Session summary data files

These files contains summary data from the individual experimental sessions, as tab delimited text with a header row that provides column labels. The columns are as follows

• subject - subject initials
• experiment - pointer to condition file label (meaningful only within the Brainard lab)
• condition - condition number within condition file (meaningful only within the Brainard lab)
• session - sequential session number for this condition
• targetLum - nominal luminance of test object being judged (cd/m2)
• rStairMean - mean r chromaticity of staircases that converged
• bStairMean - mean b chromaticity of staircases that converged
• rStairStd - r chromaticity standard deviation of staircase pairs that converged
• bStairStd - b chromaticitiy standard deviation of staircase pairs that converged
• meanu - Nominal CIE u chromaticity of test object achromatic point. Computed from mean staircase r and b coordinates. Does not take tone mapping done after test synthesis into account, and the test object locations were not indicated as precisely as in the final post-experimental processing. For further analysis, we use the achromFromPixels extracted from the image corresponding to the mean achromatic r and b chromaticities. These values that are provided as columns further along in this file.
• meanv - Nominal CIE v chromaticity of test object achromatic point
• nGoodStairs - number of converged staircase pairs within the session
• reversalsBad - 1 means not enough staircase pairs converged within the session, 0 means that enough did converge.
• stdBad - 1 means that the within session r/b standard deviation was too large to use this session's data, 0 means session OK by this criterion
• ciRefCondition - which condition number provides the reference data to compute constancy index for the data in this row
• light - illuminant label
• background - background condition label
• shape - test object shape label
• material - test object material
• illumu - CIE u chromaticity of illuminant
• illumv - CIE v chromaticity of illuminant
• testBackgroundu - CIE u chromaticity of the immediate surround of the test object
• testBackgroundv - CIE v chromaticity of the immediate surround of the test object
• testBackgroundY - Luminance (cd/m2) of the immediate surround of the test object
• achromFromPixelsu - CIE u chromaticity of achromatic point obtained by averging test object pixels in displayed image. This done by constructing the image corresponding to the mean within session staircase r and b chromaticities and then taking the mean XYZ coordinates of the pixels corresponding to the test object. This mean then converted to uvY.
• achromFromPixelsv - CIE v chromaticity of achromatic point obtained by averging test object pixels in displayed image
• achromFromPixelsY - Luminance (cd/m2) of achromatic point obtained by averaging test objects pixels in displayed image

Condition summary data files

These files contain summary data from the individual experimental conditions, as tab delimited text with a header row that provides column labels. It was obtained by processing the session summary files. The columns are as follows

• subject - subject initials
• experiment - pointer to condition file label (meaningful only within the Brainard lab)
• condition - condition number within condition file (meaningful only within the Brainard lab)
• targetLum - nominal luminance of test object being judged (cd/m2)
• ciRefCondition - which condition number provides the reference data to compute constancy index for the data in this row
• light - illuminant label
• background - background condition label
• shape - test object shape label
• material - test object material
• contextSpheres - context sphere material label
• illumu - CIE u chromaticity of illuminant
• illumv - CIE v chromaticity of illuminant
• testBackgroundu - CIE u chromaticity of the immediate surround of the test object
• testBackgroundv - CIE v chromaticity of the immediate surround of the test object
• testBackgroundY - Luminance (cd/m2) of the immediate surround of the test object
• nGoodSessions - Number of good sessions for this condition
• sessionIndex1 - Index number of first good session
• achromFromPixelsu1 - CIE u chromaticity of achromatic point obtained by averging test object pixels in displayed image.
• achromFromPixelsv1 - CIE v chromaticity of achromatic point obtained by averging test object pixels in displayed image
• achromFromPixelsY1 - Luminance (cd/m2) of achromatic point obtained by averaging test objects pixels in displayed image
  • Quadruples of coloums of the same form as the four above then follow, one set for each good session. To keep the tab delimited format aligned, values of -1 are entered for conditions where there are fewer good sessions than the maximum number obtained for any session.
• grandUMean - Mean CIE u chromaticity of achromatic point, obtained over sessions from the data in the preceding columns. Mean obtained from chromaticities, not trisimulus values.
• grandUMean - Mean CIE v chromaticity of achromatic point, obtained over sessions from the data in the preceding columns. Mean obtained from chromaticities, not trisimulus values.
• grandYMean - Mean luminance (cd/m2) of achromatic point, obtained over sessions from the data in the preceding columns
• steUMean - standard error of grandUMean
• steVMean - standard error of grandVMean
• CI - constancy index. -1 means that it could not be computed. This is either because the row corresponds to a reference condition or because there were no good data for the reference condition.
• steCI - standard error of the constancy index. -1 means that it was not computed. To put an error on the constancy indices, one has to decide how to pair up conditions across sessions. This would be possible with a boostrapping method, but we are not currently (June, 2011) doing so, because we have focussed our statistical analyses on between observer comparisons. The column is left to make it easy to include this later if we ever decide we really want it.
• eqIllumu - CIE u chromaticity of equivalent illuminant. -1 means that it could not be computed for same possible reasons as constancy indices.
• eqIllumv - CIE v chromaticity of equivalent illuminant. -1 means that it could not be computed for same possible reasons as constancy indices.