Color Constancy in a Naturalistic Goal-Directed Task

Supplemental Material

Ana Radonjić, Nicolas P. Cottaris & David H. Brainard


Experimental instructions

Instructions (pdf)
Note: Experimental intstructions are slighly modified after Experiment 1. Sections of instructions added at this point are marked in italics.    

Stimulus geometry

Blender files specifying the position of each element in each of the stimulus scenes:
(NOTE: Blender is required to view these files).

Experiments 1 and 2: Model/Workspace        Source
Experiment 3: Model/Workspace        Source

Stimulus colorimetry

Illuminants used for rendering. Spectral power distribution for each simulated illuminant as tab delimited text.
Table format: 1st column: wavelength in nm (380 - 780 nm, 5 nm steps); 2nd column: illuminant spectrum.
Different spectral sampling was used for rendering (400 - 700 nm, 10 nm steps).
Experiments 1 and 2-Set A: Standard(6500K), Bluish(12000K)
Experiments 3 and 2-Set B: Standard(6500K), Yellowish(4500K), Bluish(12000K)

Illuminant spectra impinging on the blocks.
These values are extracted from images in which the blocks were rendered as perfectly reflective surfaces.
Following the methods described in the paper, we used these illuminant values to: (1) compute the reflectance of competitors (2nd block from the left served as the standard), (2) compute the reflectances of blocks in basis images (mean across blocks used as the standard) and (3) compute the desired RGB values for a chosen block of certain reflectance that is moved to a certain block location in the workspace (these values are used as input for synthesizing the chosen block in the workspace).

Table format: 1st column: wavelength in nm (380 - 780 nm, 5 nm steps); Columns 2-5: illuminant spectrum impinging on each of the four blocks (from leftmost in column 2 to rightmost in column 4).

Experiments 1 and 2-Set A: Standard(6500K), Bluish(12000K)
Experiment 2-Set B: Standard(6500K), Yellowish(4500K), Bluish(12000K)
Experiments 3: Standard(6500K), Bluish(12000K)

Targets. Simulated surface reflectance as tab delimited text.
Table format: 1st column: wavelength in nm (400 - 700 nm, 10 nm steps); 2nd column: spectral reflectance. 
The same spectral sampling was used for rendering. 

Experiments 1 and 2-SetA: Gray (T1), Rose (T2), Teal (T3), Green (T4).
Experiment 2-Set B: Corn (T1), Skin (Caucasian, T2), Lettuce (T3), Tree bark (T4).
Experiment 3: Straw(T1). T2-T4 were the same as in Experiment 2-SetB.

Competitor sets. Simulated surface reflectance for competitor is provided for each experiment and condition as tab delimited text.
Table format: 1st column: wavelength in nm (400 - 700 nm, 10 nm steps). Remaining columns: spectral reflectance for each competitor (competitor  labels are denoted in the first row).
Competitor notation is the same as that used in Figure 2.
Experiments 1 and 2-Set A: Illuminant-constant, Illuminant-change (bluish).                             
Experiments 2-Set B: Illuminant-constant, Illuminant-change (yellowish), Illuminant-change (bluish)
Experiment 3: Illuminant-constant, Illuminant-change (bluish)
CIELAB and xyY values of each target and its competitor set are tabulated for each illumination condition. Competitor notation is the same as that used in Figure 2. For each set we also provide a white point used for conversion.
Experiments 1 and 2-Set A: Target and competitor blocks (CIELAB and xyY)                             
Experiments 2-Set B: Target and competitor blocks (CIELAB and xyY)
Experiment 3: Target and competitor blocks (CIELAB and xyY)
CIELAB Distance table provides distances in (delta E) between adjacent competitors in a set for all experiments and conditions (pdf).

Individual Observer Data

Experiment 1: DLV, NIT, KJF, CTG.
Experiment 2A: LTD, IEA, HGC, LDO, EAZ, VSR.
Experiment 2B: LTD, IEA, LDO, EAZ, VSR.
Experiment 3: SED, MGO, ITY, BSW, NKJ, GTH.

Preliminary experiment for Experiment 3 - methods and results (pdf)