**
Project title: Colour and colour vision with Ostwald, device, and elementary colours -
Antagonistic colour-vision models TUBJND and TUBLAB, and properties for many applications
Chapter E: Colour Metrics, Differences, and Appearance (2023),
Main part eea_s, under work**

eea_s41 in English or ega_s41 in German.

eea_s42 in English or ega_s42 in German.

eea_s43 in English or ega_s43 in German.

eea_s44 in English or ega_s44 in German.

eea_s45 in English or ega_s45 in German.

eea_s46 in English or ega_s46 in German.

For links to the main chapter E

Content list of chapter E (links and file names use small letters), see

eea_s in English or ega_s in German.

_________________________

**Title of part 4.1: Achromatic colour metric for the surface colour range for adjacent (a) and separated (s) samples**

The term "antagonistic" of the Greek language is well known in medicine for muscles. The muscles may work in "antagonistic" (or opponent) directions.

Colours may have "antagonistic" properties, which for example can be described by "opponent", "complementary", "compensatory", "shadow effect", "after image effect", "successive contrast", and "simultan contrast".

The "antagonistic" colour attributes can be described by

"lightness - darkness", "blackness - brilliantness", "whiteness - deepness", "chromaticness - achromaticness".

The question arises, if an "antagonistic" colorimetry is able to describe many colour effects and colour attributes. These affects and attributes are applied in the area of design, architecture, and art, and for example in the Swedish

The ISO/CIE colorimetry, for example of CIELAB, CIELUV, and CIEDE2000 according to ISO/CIE 11664-2, 5, and 4, is mainly used for surface colours. The tristimulus-value range may be

The ratio

The following TUBJND colorimetry is intended for a much larger surround-luminance range 10 <=

The colour spaces CIELAB and CIELUV include a colour difference formula. However, CIEDE2000 is a colour difference formula without a colour space. This produces problems for many applications.

Similar LABJND according to CIE 230:2019

The members of the CIE committee TC1-81, who developed CIE 230, have discussed a

This was one reason, that CIE 230 includes only a colour-difference formula and no colour space.

The information on this TUB web site is intended to present a line element for LABJND according to CIE 230 for the surface colour range. The surface colour range for mate samples is the tristimulus value range

The name of the colour difference formula LABJND 230 is changed here to TUBJNDa, and the colour space defined by a line element is called TUBLABa. The index a indicates application for adjacent (a) colours, similar as for CIEDE2000.

It is important to know, that the colours spaces CIELAB and CIELUV have been developed for separate (s) samples on a grey surround. There is a fundamental metric difference between adjacent and separate colours. The slope in appropriate figures may differ by up to a factor 2, see

Users may view samples of the luminance

For the download of this figure in the VG-PDF format, see eea00-1n.pdf.

For the colorimetric description of the samples and the surround one can use the reflection factor

For the download of this figure in the VG-PDF format, see eea00-2n.pdf.

The samples could be viewed adjacent (a) or separate (s). The surround may vary between black, grey or white, see Fig. 4.11-3.

For the download of this figure in the VG-PDF format, see eea00-3n.pdf.

For separate (s) samples on the surrounds black, grey or white different scaling functions for the lightness

For the download of this figure in the VG-PDF format, see eea00-4n.pdf.

CIELAB is determined by separate colours.

TUBJND is determined by adjacent colours.

Assumption:

The colour difference is dependent on the

1. two adjacent samples

2. the sample and the surround

For the tristimulus value difference (d) compared to the surround one calculates in the two cases because of

1.

2.

If it is agreed that the lightness

Proof:

For the grey surround the lightness

For the colour difference of LABJNDa according to CIE 230 the slope is between 1,09 and 1,16. Therefore the slope is about twice of LABJNDa compared to CIELABs.

This proof seem to confirm the assumption, that the mean tristimulus value

In my view there may several other conflicts with the results of CIE TC1-81. CIEDE2000 is the only CIE formula intended to describe colour differences of adjacent colours. However, CIEDE2000 has no colour space, and CIELAB is intended for separate colour samples.

In the scope range 0 <=

In addition the performance of LABJND seems similar and not worse compared to CIEDE2000 for Large Colour Differences (LCD). The performance of 6 datasets for Large Colour Differences (LCD) in the range 5 <=

The colour-difference formula LABJND of CIE 230:2019 and new research results for a larger luminance range may be a good starting point for colour-difference equations and colour spaces for a luminance range of about 5 log units including the surface-colour luminance range of about 1,5 log units.

New research results of

Physiological and psychophysical data for achromatic colours for the luminance range of about 5 log units compared to 1,5 log units for surface colours allow to extend the above results to achromatic colours for the luminance range of about 5 log units, see part 4.3.

Finally some assumptions about the properties of the

The TUB colour vision model for the luminance range of about 5 log units may improve the colour-difference formula and colour space CIELABa for the surface-colour luminance range. In addition the colour-appearance attributes of the TUB colour-vision model for about 5 log units may be very usefull for image applications in the High Dynamic Range (HDR).

For example

https://aic-color.org/publications,

go to AIC 1971 to download the paper of

For the download of this figure in the VG-PDF format, see een31-1n.pdf.

For

For the download of this figure in the VG-PDF format, see eeo00-4n.pdf.

The constants of the line element equations and the experimental data will be described further in the section 4.14. The line element uses for all colours the tristimulus value

For

For the download of this figure in the VG-PDF format, see eer30-2n.pdf.

For the tristimulus value of white

If instead of

For the download of this figure in the VG-PDF format, see eer31-4n.pdf.

Visually the sRGB output appears equally spaced on displays, if the relative tristimulus values of the three primaries are used to calculate the lightness

According to the ergonomic standard ISO 9241-306:2018 for display-work places, the equally spaced

For achromatic surface colours the tristimulus value

The question raises how the blue primary looks for the tristimulus value range 1,3 <=

In addition for the triangle contrast

For the download of this figure in the VG-PDF format, see eer31-7n.pdf.

If the CIELAB lightness

For the download of this figure in the VG-PDF format, see eer31-3n.pdf.

For the colour scales between black N and W, G, R, or B the CIELAB lightness

At display work places the reflection of the ambient light changes all colours by a mixture with the achromatic reflection W. For example the 4% standard reflection compared to the white display reduces the colour gamut to about 70%, see ISO CEN DIN 9241-306:2018/9.

The three primaries

The

The Just Noticeable Difference (JND) has been vsually determined by six observers for two adjacent samples in a grey surround with a white border, see Fig. 4.13-1.

For the download of this figure in the VG-PDF format, see eej11-3n.pdf.

The figure shows a linear equation between

For the download of this figure in the VG-PDF format, see eeo10-2n.pdf.

Equations [5d] and [6d] are linear equations. Equations [7d] and [8d] are potential (nonlinear) equations. A value for the least square error is given. The accuracy for the potential approximation is higher. In both cases the calculation of a line element is possible by integration. One can expect an increase of the accuracy by a factor 2,5 by the potential approximation in equation [8i] compared to the linear equation in equation [6i].

One must consider that the luminance range of surface colours is only 1,5 log units compared to the about 5 units of the HDR luminance range. In chapter 3 we will use exponential (nonlinear) approximations to describe

A linear and potential line element connection between TUBJNDa and TUBLABa is given in the last Fig. 4.13-2. In both cases the constant A1 will be obsolete, if the

For the download of this figure in the VG-PDF format, see eej61-5n.pdf.

In addition the constant A1 will be obsolet, if the

For comparison of experimental results usually the

The experimental data of the BAM-results are shown for the

1. lightness

2. tristimulus value difference at threshold

3. sensitivity

4. contrast

on a linear ordinate scale, see Fig. 4.14-2.

For the download of this figure in the VG-PDF format, see eej80-3n.pdf.

The experimental data of the BAM-results are shown for the

1. lightness

2. tristimulus value difference at threshold

3. sensitivity

4. contrast

on a log ordinate scale Fig. 4.14-3.

For the download of this figure in the VG-PDF format, see eej60-3n.pdf.

On a log scale the nonlinear dependance of some of the four relative values may show a linear part in the figure, see for example the contrast for

In the following the

For the download of this figure in the VG-PDF format, see eej71-5n.pdf.

The experimental data of

1. lightness

2. tristimulus value difference at threshold

3. sensitivity

4. contrast

on a linear ordinate scale see Fig. 4.15-2.

For the download of this figure in the VG-PDF format, see eej90-3n.pdf.

The experimental data of

1. lightness

2. tristimulus value difference at threshold

3. sensitivity

4. contrast

on a log ordinate scale, see Fig. 4.15-3.

For the download of this figure in the VG-PDF format, see eej70-3n.pdf.

On a log scale the nonlinear dependance of some of the four relative values may show a linear part in the log-log figure, see for example the contrast for all

The log ordinate of Figure 4.15-3 is more of theoretical value. One must consider that only the linear ordinate in Fig. 4.15-2 relates to visual properties in applications.

CIEDE2000 is a colour-difference formula and has no colour space. Therefore the color space coordinate lightness is not defined. There is a larger complexity with two separate parts of the CIEDE2000 formula. The colour difference formula includes the lightness difference

Therefore, it seems

For the comparison of the lightness

For the download of this figure in the VG-PDF format, see CET51-3N.PDF.

The comparison of the contrast shows similarities for LABJND and CIEDE2000 for all lighter colours compared to the grey surround.

However, the differences are larger on a linear ordinate scale, which corresponds to the visual evaluation. For the relative lightness

For the download of this figure in the VG-PDF format, see eef20-1a.pdf.

The value of the

For the download of this figure in the VG-PDF format, see eea00-3n.pdf.

However in the viewing situations adjacent of Fig. 4.16-3, the lightness in Fig. 4.16-2 is smaller for LABJND compared to the separate samples of CIELAB. However, the difference

For the relative contrast

For the download of thie three figures in the VG-PDF format, see eew40-3n.pdf

The log ordinate of Figure 4.16-4 is more of theoretical value (still missing). One must consider that only the linear ordinate in this Figure relates to visual properties in applications.

A reason or explanation for the large contrast difference between CIELAB and LABJND is unknown. The vieving of separate samples according to CIELAB and the adjacent samples of LABJND may be the reason as shown in the dissertation of

Both LABJND and CIEDE2000 are based on viewing of adjacent samples. Therefore LABJND and CIEDE2000 may produce the more similar contrast, at least for all samples lighter compared to the surround

For the download of the two figures in the VG-PDF format, see eer40-8a.pdf and eer50-8a.pdf.

Figure 4.16-5 shows the LABJND contrast for the linear approximation of the experimental BAM data. If the potential approximation of the experimental BAM data is used, then a negative slope appears for all

For this topic the following references may be important:

Forschungsbericht 115, Bundesanstalt für Materialforschung und -prüfung (BAM), (Research report 115, Federal Institute for Material Research and Testing (BAM)), 1985, 119 pages, 87 figures, 50 tables, 184 MB, ISSN 0938-5533, ISBN 3-88314-420-7, "Open Access" since 2015, see

https://nbn-resolving.org/urn:nbn:de:kobv:b43-3350

CIE 230:2019,

CIE 230 has been developed by CIE TC1-81 with the chairman:

http://www.cie.co.at/publications/validity-formulae-predicting-small-colour-differences .

CIE 230 has used CIE datasets with about 7000 pairs of adjacent colour samples from different countries, see

http://files.cie.co.at/TC181_Datasets.zip.

The files are provided in three different formats: .TXT (text), .PS (PostScript), and .PDF (Portable Document).

CIE 230 has studied the validity of five colour difference formulae. For example the performance of the formula LABJND 1985 for adjacent colours is compared with the formula CIELAB for separate colours on a grey background. LABJND uses the logarithmic

For example out of 8 CIE datasets, and for the intended colour differences < 2 CIELAB of CIE 230, the formula LABJND_PF performs best in 5 cases, and CIEDE2000_PF in 1 case. For example out of 13 (all) CIE datasets, and for the colour differences < 5 CIELAB, the formula CIEDE2000_PF performs best in 6, and LABJND_PF in 5 cases. CIELAB is best in 1 case, see section "7 Conclusions" of CIE 230. According to this section E, part 1, one may consider the advantage of a lightness-line element in LABJND which is missing in CIEDE2000.

The performance for large colour differences of LABJND and CIEDE2000 have been studied in CIE TC1-63. The CIE requirement of a complete agreement of the TC members on a CIE report was not possible. Therefore TC1-63 was deleted in 2019 without a CIE report.

For example the calculated performance for Large Colour Differences (LCD) show

For the download of this figure in the VG-PDF format, see YE370-7N.PDF.

For the colour-difference range of LCD colours in the range 5 <=

It is intended to repeat the calculations of this Fig. 4.17-1 of 2016 with the new research results of this chapter E, compare results of the following parts 4.2 to 4.4.

_________________________

For links to the main chapter E

Content list of chapter E (links and file names use small letters), see

eea_s in English or ega_s in German.

In future the figures and the text may be improved.

Link to the next topic (under work in 2023)

eea_s42 in English or ega_s42 in German.

-------

For the TUB start site (not archive), see

index.html in English, or
indexDE.html in German.

For the TUB archive site (2000-2009) of the BAM server
"www.ps.bam.de" (2000-2018)

about colour test charts, colorimetric calculations, standards,
and publications, see

indexAE.html in English,
indexAG.html in German.

For similar Information of the BAM server "www.ps.bam.de"
from the WBM server (WayBackMachine)

https://web.archive.org/web/20090402212108/http://www.ps.bam.de/index.html