Cone Array Models

and

Color Calibration

Al Ahumada

NASA Ames Research Center

Outline

Cone array

Density and Distribution

Perry & Cowey (VR 1985) Monkey complete lattice

Curcio, et al. (Sci 1987) Complete human lattice

Models

Ahumada & Poirson (JOSA 1987) Sequential disk

Sloan & Curcio (SPIE 1991) [Curcio & Sloan (VN 1992)] Jittered lattice

Psychophysics

Live Human Imaging

Artal & Navarro (OL 1989) Spatial power spectrum

Miller et al. (VR 1996) Direct image

S Cone Array Density and Distribution

de Monasterio, et al. (IOVS 1985) Primate array and model

Ahnelt, Kolb, Pflug (JCN 1987) Human array identified by anatomical features

Curcio, et al. (JCN 1991) Human array identified by staining

L-M Cone Array

Ratio

Vimal et al. (VR 1989) 1.7:1, 4:1 based on spot detection in 2 Os

Bowmaker (NRC 1990) Microspectrophotometry: human 1.67:1 (164:98), monkey 0.77:1 (101:131)

Regularity

Mullen & Kingdom (ARVO 1991) Human discrimination with eccentricity consistent with random array and indiscriminate assignment

Direct Evidence Mollon & Bowmaker (Nature 1992). Five monkey fovea fragments. Mathematica analysis shows same-neighbor prob. = 0.53 (not significantly non-random [but close when arrays randomized separately]).

Packer, Williams, & Bensinger (JN 1996) Monkey periphery same-neighbor prob. = 0.65 (significantly non-random, p<0.006)

Suggestions: Fall in discriminability with eccentricity may relate to "clumping". Other patterns should be considered: Boon & Noullez (1986)

Color Learning and Calibration Algorithms

Learnability of L-M opponency from L & M cone responses:

Cone specific responses in the LGN:

Reid & Shapley (1992) M sequence results show opponent centers in LGN

Lee & Kremers (1997) Edge stimuli show specific, rather than mixed, surrounds

Model incorrect for learning cortical color system from LGN.

Non-specific wiring of horizontal cells

Model may account for learning of the cone specific LGN responses.

Need for translation-invariant color calibration