NASA’s Tricky, Trippy Games With the Color Spectrum

[bystander]

NASA’s Tricky, Trippy Games With the Color Spectrum
Discover Blogs | Visual Science | 20 May 2010

Though humans cannot see light outside the visible spectrum, satellites are able to detect wavelengths into the ultraviolet and infrared. The Landsat 7 satellite uses an instrument that collects seven images at once, with each image showing a specific section of the electromagnetic spectrum, called a band. Each image highlights a different part of the electromagnetic spectrum.

The satellites original images are all acquired in black and white, so color must be assigned via computer to the black and white images. The three primary colors of light are red, green and blue, and each color is given a different band/image. Once the three images are combined, you will have what is called a “false color image.” A common band combination shows green, healthy vegetation as bright red, which is useful in forestry and agricultural applications. Landsat images are also used to gather geological and hydrological data along with other kinds of environmental monitoring. In a helpful explanation, the folks at Landsat offer this catchy formula to aid you in remembering: “One common way that primary colors are assigned to bands can be easily remembered using the mnemonic:

“RGB = NRG (Red, Green, Blue = Near Infrared, Red, Green…)
Red = Near IR (ETM+ band 4)
Green = Red (ETM+ band 3)
Blue = Green (ETM+ band 2)”

There. That should be easy to remember. My suggestion: try singing it.

All images courtesy USGS National Center for EROS and NASA Landsat Project Science Office

Ganges River Delta. The Ganges River forms an extensive delta where it empties into the Bay of Bengal.


Kamchatka Peninsula. The eastern side of Russia’s Kamchatka Peninsula juts into the Pacific Ocean west of Alaska.


Boliva, Amazon Basin. This image of Bolivia shows dramatic deforestation in the Amazon Basin. Loggers have cut long
paths into the forest, while ranchers have cleared large blocks for their herds. Fanning out from these clear-cut areas
are settlements built in radial arrangements of fields and farms. Healthy vegetation appears bright red in this image.


The Himalayas. The snow-capped peaks and ridges of the eastern Himalayas Mountains create an irregular
white-on-red patchwork between major rivers in southwestern China.


Terkezi Oasis. A series of rocky outcroppings are a prominent feature of this Sahara Desert landscape near the
Terkezi Oasis in the country of Chad.

[biddie67]

....... The satellites original images are all acquired in black and white, so color must be assigned via computer to the black and white images. The three primary colors of light are red, green and blue, and each color is given a different band/image. Once the three images are combined, you will have what is called a “false color image.” A common band combination shows green, healthy vegetation as bright red, which is useful in forestry and agricultural applications. ........... the folks at Landsat offer this catchy formula to aid you in remembering: “One common way that primary colors are assigned to bands can be easily remembered using the mnemonic:

“RGB = NRG (Red, Green, Blue = Near Infrared, Red, Green…)
Red = Near IR (ETM+ band 4)
Green = Red (ETM+ band 3)
Blue = Green (ETM+ band 2)”

....

From my untrained eye, their color substitution choices don't make any sense - why not assign a 4th color for IR and keep RGB as RGB?

[RJN]

http://antwrp.gsfc.nasa.gov/apod/ap021112.html

[bystander]

Should have been a SMO. I would be looking for HiRISE pictures of Mars.

[Chris Peterson]

From my untrained eye, their color substitution choices don't make any sense - why not assign a 4th color for IR and keep RGB as RGB?

There is no fourth color. Ultimately, the multispectral data has to be assigned to the three available primaries, which are red, green, and blue for most display devices (and approximately that for the human eye). If you have three data bands, you can arbitrarily assign them to RGB according to what seems to make the data most clear, or even based on aesthetics. If you have more than three bands (as is often the case), you need to apply some sort of mathematical model, mixing ratios of different bands before assigning them to either red, green, or blue. More complex models may mix input bands and then split them into output mixes of red, green, and blue- something like your idea of adding a fourth color, but not exactly.

It is often desirable in images like those posted above to deliberately force the color scheme away from RGB, since a completely unnatural appearance actually makes it easier for people to see details that might otherwise not be noticed.

[biddie67]

...... It is often desirable in images like those posted above to deliberately force the color scheme away from RGB, since a completely unnatural appearance actually makes it easier for people to see details that might otherwise not be noticed.

It is interesting that an awareness of human recognition skills is figured into a best fit analysis for presentation. After all, this data was ordered by humans for humans - not aliens!! ((grin))

...... apply some sort of mathematical model, mixing ratios of different bands before assigning them to either red, green, or blue. More complex models may mix input bands and then split them into output mixes of red, green, and blue- something like your idea of adding a fourth color, but not exactly.

Is there somewhere on the internet, a very high-level explanation of the math model and mixing ratios that are used for the resultant displays? I'd be interested in getting a sense of the thought processes behind this type of modeling. I know that I wouldn't be able to understand the actual math.