This paper addresses the challenge we face when attempting to judge the quality of color reproduction obtained from digital still cameras. We will consider possible methods and metrics and show how traditional colorimetric analysis (as used in scanning technology) has limited use as a tool for judging rendered digital camera imagery. We will talk about metrics that can be used to rate colorimetric accuracy of scene analysis of a digital imaging device and the constrained situations in which they are useful. Digital still camera can be classified into two types: those that allow access to unrendered un-color-corrected data, and those that do not. The first type, most of which are professional cameras, allow someone testing the camera to measure the linearity and the spectral sensitivity of the camera system and determine metrics that indicate how well the camera can see color. This type of analysis can also be applied directly to the imager (such as a charged-coupled device) with or without IR filters. Although this analysis does not consider the resultant rendered image, it is the only reliable method yet proposed to compare colorimetric capabilities of digital cameras. The second camera type (and the first type after rendering is applied to the image data) requires a subjective analysis of a range of scenes under extremely different circumstances. A discussion and several examples of what is involved in white-point balancing of digital images justifies our claim that color metrics are unsuited for judging rendered color quality in digital cameras. We describe how color appearance models fall short in their determination of the adapted white point of a scene and are thus unreliable as metrics for color image quality in digital cameras. The color transformations and the tonal rendering that are applied to camera images for display are also described briefly in terms of their effect on color image quality. We present a recommendation for types of scenes that could give efficient testing of color quality when used in a subjective analysis.
[1]
Mark S. Drew,et al.
The Maximum Ignorance Assumption with Positivity
,
1996,
Color Imaging Conference.
[2]
Michael H. Brill,et al.
Color appearance models
,
1998
.
[3]
M. R. Pointer,et al.
Measuring Colour Reproduction
,
1986
.
[4]
Graham D. Finlayson,et al.
Recovering Device Sensitivities with Quadratic Programming
,
1998,
Color Imaging Conference.
[5]
Robert G. Gann.
Desktop Scanners: Image Quality Evaluation
,
1998
.
[6]
Michael H. Brill,et al.
Prime colours and Colour Imaging
,
1998
.
[7]
Graham D. Finlayson,et al.
Illuminant estimation and colour correction
,
1998
.
[8]
Michael H. Brill,et al.
Prime colors and color imaging
,
1998,
Color Imaging Conference.
[9]
Mark D. Fairchild,et al.
Color Appearance Models
,
1997,
Computer Vision, A Reference Guide.
[10]
Mark S. Drew,et al.
Matrix Calculations for Digital Photography
,
1997,
Color Imaging Conference.
[11]
Andrew K. Juenger.
Color Sensitivity Selection for Electronic StilI Cameras Based on Noise-Considerations in Photographic Speed Maximization
,
1998,
PICS.
[12]
R. Hunt.
Revised colour‐appearance model for related and unrelated colours
,
1991
.
[13]
Jack M. Holm.
Adjusting for the Scene Adopted White
,
1999,
PICS.
[14]
Graham D. Finlayson,et al.
Color by Correlation
,
1997,
CIC.
[15]
W. A. Thornton,et al.
Spectral sensitivities of the normal human visual system, color‐matching functions and their principles, and how and why the two sets should coincide
,
1999
.
[16]
G. K. Smelser.
The structure of the eye
,
1961
.