Luminance/Colour Images

article by Ben Lincoln

Unlike most of the other articles in this section, this one describes a technique that seems to be mostly used by photographers and artists than scientists and other technical people. I am not sure why this is - it seems to be extremely useful for combining different aspects of multispectral imagery.

The technique in question is one in which the colour of one image is applied to the luminance (perceived brightness^[1]) of another. I suspect that one of the reasons it may be used mostly for artistic purposes is that there is no standard method of performing this operation. "Luminance" is a fairly well-recognized term, although there are two different systems that are used to calculate it^[2]. "Colour" is a more ambiguous term, I suppose.

In any case, I think the results are often extremely revealing, and there is a lot of unexplored potential in this area.

The specific method I use is a variation on the one used by the "Colour" blend mode in Adobe Photoshop®. See Secrets of Photoshop's Colour Blend Mode Revealed (Sort Of) for more detail than you probably care to know, as well as comparisons against several other options.

Basic Use

When I first started using this technique, the two main sources that I'd use for luminance were the near infrared and ultraviolet-A versions of a shot. The main sources for colour were the red/green/blue variation, and the tinted greyscale version of whichever of near infrared and ultraviolet-A was not being used for the luminance channel. Near infrared in particular tends to give an otherworldly appearance to the image when used in this way.

Luminance/Colour Examples 1

Oregon Coast

Ajo Mountains

Saguaro National Park

West Thumb Geyser Basin Pool

Crater Lake

In this set of example images, the luminance of the near infrared version of each image has been combined with the colour of the red/green/blue version.

Luminance/Colour Examples 2

Ajo Mountains

Crater Lake

West Thumb Geyser Basin Pool

In this set of example images, the luminance of the near infrared version of each image has been combined with the colour of a tinted version of a second greyscale channel from the same image. The pink and purple tones are from tinted ultraviolet-A variations, and the green tone is from a tinted greyscale image based on the luminance of the red/green/blue version of that image.

Andrea G. Blum also has a few images in her lovely gallery of UV flower photos that were made in this manner, using ultraviolet-A as the luminance channel.

Complex Use

With the statistical images made available by The Mirror's Surface Breaks, I find that in addition to near infrared, I am usually happy with the results obtained by using the average deviation (or standard deviation - they're pretty close in most cases) and/or variance as the luminance channel of a colour image. Kurtosis and skewness often provide interesting results as well.

Luminance/Colour Examples 3

ADev Luminance / (G-B-UVA) Colour [LC]

Kurtosis Luminance / (R-G-B) Colour [LC]

Kurtosis Luminance / ((Ratio: G To NIR)-(Ratio: B To NIR)-(Ratio: UVA To NIR)) Colour [LC]

Kurtosis Luminance / ((Ratio: R To NIR)-(Ratio: G To R)-(Ratio: UVA To B)) Colour [LC]

Skewness Luminance / (FC-3C-NIR-G-B) Colour [LC]

Skewness Luminance / ((Ratio: G To NIR)-(Ratio: B To R)-(Ratio: UVA To B)) Colour [LC]

Various luminance/colour combinations of a macro shot of flowers.

Date Shot: 2010-10-16
Camera Body: Nikon D70 (Modified)
Lens: Nikon Micro-Nikkor 105mm f/4
Filters: Standard Set
Date Processed: 2011-01-16
Version: 1.0

Luminance/Colour Examples 4

Kurtosis Luminance / (NIR-B-UVA) Colour [L/C]

Kurtosis Luminance / ((Ratio: G To NIR)-(Ratio: B To NIR)-(Ratio: UVA To NIR)) Colour [LC]

Skew Luminance / ((Ratio: R To UVA)-(Ratio: G To UVA)-(Ratio: B To UVA)) Colour [LC]

A few examples using kurtosis and skewness as the luminance channel, based on an image of Santa Elena Canyon in Big Bend National Park.

Certain applications of this technique can give surprising results. Compare the red/green/blue and luminance/colour versions of these two photos of a grove of orange trees in Florida, and notice how brightly the fruit stands out in the latter. I don't know how reliably this works, but it seems to me that this would be much easier for automated fruit-harvesting systems to recognize than the conventional image.

Luminance/Colour Examples 5

R-G-B

NIR Luminance / ((ND (Pos): R-G)-(ND (Pos): R-B)-(ND (Pos): R-UVA)) Colour

R-G-B

NIR Luminance / ((ND (Pos): R-G)-(ND (Pos): R-B)-(ND (Pos): R-UVA)) Colour

A somewhat complex luminance/colour combination makes the oranges on these trees stand out quite prominently from the other elements.

This method of using on of the statistical images for luminance can produce maps from satellite imagery that combine the utility of certain false colour representations with the higher contrast/viewability of other variations. In the following examples, note in particular how the vegetation index-based images of New York City could easily be used as a map to the areas there that have a large amount of plant life.

Luminance/Colour Examples 6

ADev Luminance / (FC-Gr-(ARVI (Full Range))-Vegetation Index Dual Scale) Colour [LC]

ADev Luminance / (FC-Gr-(ARVI (Full Range))-Vegetation Index Single Scale) Colour [LC]

ADev Luminance / (FC-Gr-(GNDVI (Full Range))-Vegetation Index Triple Scale 4) Colour [LC]

Several variations on satellite imagery of Big Bend National Park.

Luminance/Colour Examples 7

IR1 Luminance / (FC-3C-IR3-Red-Green) Colour [LC]

IR1 Luminance / (FC-3C-SRTM-ThH-ThL) Colour [LC]

SRTM Luminance / (FC-3C-ThH-IR1-Green) Colour [LC]

SRTM Luminance / (FC-3C-ThH-IR3-IR1) Colour [LC]

Several variations on satellite imagery of Egypt, near Luxor.

Luminance/Colour Examples 8

Average Deviation Luminance / ((GEMI (Full Range))-Vegetation Index Dual Scale) Colour [LC]

Average Deviation Luminance / ((NDVI (Full Range))-Vegetation Index Single Scale) Colour [LC]

Variance Luminance / ((GEMI (Full Range))-Vegetation Index Dual Scale) Colour [LC]

Skewness Luminance / ((ND53 (Full Range))-Vegetation Index Triple Scale 4) Colour [LC]

Skewness Luminance / ((Ratio: IR2 To Green)-Modern Thermograph 2) Colour [LC]

Red/Green/Blue (Luma) Luminance / (FC-3C-SRTM-ThH-IR2) Colour [LC]

Red/Green/Blue (Luma) Luminance / (FC-3C-ThL-IR3-IR1) Colour [LC]

Skewness Luminance / (ThH-IR1-Red/Green/Blue (Luma)) Colour [LC]

Average Deviation Luminance / ((ThL)-Classic Thermograph) Colour [LC]

Average Deviation Luminance-((ThL)-Modern Thermograph 3) Colour [LC]

Several variations on satellite imagery of New York City.

Tinting

Although there are numerous methods of tinting greyscale images, the one I use is to apply a solid colour (instead of a separate colour image) using this same luma/colour technique. There are examples of this in nearly every photo set on this site.

Footnotes

"Luminance" is usually abbreviated as "Y" in the names of colourspaces which make use of it. This is because the concept of luminance originated with the CIE's "XYZ" colourspace, and I assume there was named "Y" simply because the three most common mathematical names for variables were chosen.
Historically, I've tended to use the terms "luminance" and "luma" interchangeably, although I try to remember to spell out "luminance" now, because there is technically a difference between them. There are a couple of reasons for this:

"Luma" is shorter (for filenames, captions, and other space-constrained uses), but more descriptive than "Y", and a lot less ambiguous than "L". This is the main reason you will still find me using "luma" in the broader "perceived brightness" sense.
Even though people like Charles Poynton claim that there was no use of the term "luma" prior to their definition (I don't have time to research whether this is accurate or not), there most certainly were uses of the term "chroma" as an abbreviation for "chrominance". Since luminance and chrominance go hand-in-hand, it makes sense that "luma" would be treated as an abbreviation of "luminance".
Even in an ideal scenario where there is no other ambiguity, they're basically the same thing. The distinction is something of concern only to engineers and developers building video systems. I work in fairly technical ways with images, and I still don't care whether a channel has been gamma-compressed or not, as long as the underlying system maintains consistency in that respect between all channels.
The people who developed them picked nearly identical names and abbreviations for them, and the "correct" abbreviation for "luma" requires the use of a superscript 1, which is usually hard to include in text even in systems where superscript is supported.
Not everyone agrees on the terms, and so "Y" in an abbreviation (such as "HSY") is equally likely to mean "luma" as "luminance".

This is disregarding systems like Photoshop®, which use "luminance" to mean "the very specific method of calculating perceived brightness that Adobe's engineers developed and which no one else uses" (see Secrets of Photoshop's Colour Blend Mode Revealed (Sort Of)).
Essentially, expecting anyone other than video engineers to make a distinction between "luma" and "luminance" is the sort of losing battle that computer hackers (in the old sense) are still sometimes fighting by trying to promote the use of the term "crackers" to denote what everyone else calls "hackers", and that firearms people are still fighting by nitpicking about "magazines" versus "clips".
Note that this is completely different from people like me who try to promote the correct use of the terms "violet" versus "purple", because that distinction is much more important, and that battle will be won. Someday. By my side.

Both systems use the same formula, but assign different weights to the three primary colours. In the YUV and HSY colourspaces, luminance is calculated as being (0.587 * Green) + (0.299 * Red) + (0.114 * Blue), whereas in the YCrCb and IHSL colourspaces, it is calculated as being (0.7152 * Green) + (0.2126 * Red) + (0.0722 * Blue).