Notes to the Natural Vision (NV) data file format specification Version 2.0s

This document is slightly modified from the comments submitted to CIE TC8-07 on Oct. 26, 2003.

1. Revision

The Natural Vision (NV) data file format specification Version 2.0s is a revised version of the previous specification we provided as version 1.5s. The basic concept is not changed but some misdescriptions are corrected and detailed specifications are changed.

2. Notes on the Natural Vision data file format specification

We would like to note some points on the current NV data format specification. Although the detailed specification and the background idea are described in the specification and its' remarks we have provided already, this will make our forthcoming discussion be clear without misunderstanding.

3. Requirements to the format specification

We think the following requirements are necessary;

(1) Multichannel image
- The image data can be represented in terms of input device signal, display device signal, encoded colorimetric values, spectral reflectance (transmittance), and spectral radiance. (Output device signal for printer is the issue for future extension)
- The spectral signal, either reflectance (transmittance) or radiance can be represented by coefficients of some basis functions, such as PCA or DCT.
- The number of channels can be 3 to certain number (to be defined).
- Gray levels of each channel can be 8bit - 16bit.
- Tone curve of each channel can be either linear or nonlinear
- The background level (ex. dark current of image capturing device) can be included or not included.
- The colorimetric values can be those derived from arbitrary defined CMFs under arbitrary illuminant.
- The spectrum domain can be defined with the range 380nm-780nm at 1nm interval.
- The spectral reflectance is relative value with respect to the reflectance of a perfect diffuser (i.e., the reflectance of perfect diffuser = 1.0). The reflectance value larger than 1.0 is permitted for glossy object.
- The spectral radiance can indicate absolute values in a certain unit (e.g. w/sr/m2).

(2) Profile for input image
- The profile can be associated with the corresponding image data captured by an input device.
- The profile can be either data to provide spectral or colorimetric characterization of the image data or to estimate spectral or colorimetric data from the image data.
- The spectral sensitivity of each channel of an input device is included in the spectral characterization.
- Spectral radiance of illuminant used in image capture can be included in the spectral characterization.
- The tone curve and background level (ex. dark current of image capturing device) can be included in the characterization.
- Spectral radiance of illuminant for deriving the spectral radiance or colorimetric values of an object to be reproduced can be included in the profile.
- CMFs, either CIE 1931 XYZ or other multidimensional (>=3) extended CMFs can be included in the profile.
-
(3) Profile for display image
- The profile can be associated with the corresponding image data to be displayed by a display device.
- The profile can be either data to provide spectral or colorimetric characterization of the image data or to calculate the digital counts of the image data from spectral or colorimetric data.
- The spectral radiance of each channel of a display device is included in the spectral characterization.
- The tone curve and background spectral radiance or colorimetric values can be included in the characterization.
- CMFs, either CIE 1931 XYZ or other multidimensional (>=3) extended CMFs can be included in the profile.

(4) Profile for encoded colorimetric and spectral radiance image
- The profile can be associated with the corresponding image data representing colorimetric values and spectral radiance.
- Spectral radiance of illuminant for deriving the spectral radiance or colorimetric values of an object to be encoded can be included in the profile.
- CMFs, either CIE 1931 XYZ or other multidimensional (>=3) extended CMFs can be included in the profile.

(5) Others
- Platform independency

4. Discussion on spectral radiance or spectral stimuli

We think spectral stimuli (spectral radiance in our document) are required in the multispectral image data. There are several situations when the reflectance spectra (spectral reflectance in our document) with illuminant or colorimetric values are not enough or adequate for the extensive use. Considering the following scenes to be reproduced, the reflectance spectra can not be defined or is difficult to be obtained.

i) The scene including object colored by fluorescent material.
ii) The scene including object illuminated by plural kinds of illuminants or multiple reflection.
iii) The scene including light emitting object (e.x. the sun, ramp, etc.)
iv) The scene including light scattering object (e.x. sky, sea, glass. etc.)
v) The CG image created via three or multi-primary display device

In the fields of electric commerce, telemedicine, theater, and so on, the spectral stimuli will become necessary as the above situations.
The colorimetric values may be used in such cases, but reduce the potential of changing CMFs and spectral reproduction for discounting observer metamerism. If extended CMFs are used to derive the multidimensional colorimetric values, spectral information may be retained. From the mathematical point of view the colorimetric values derived by extended CMFs are same as spectral stimuli represented by coefficients of basis functions.


Back


(c) Natural Vision, NICT.
Any questions and bugs: please report to