False Color

Can someone help me with this? We have a nice lava lamp from the 70ies at home. The fluid is blue and the "lava" is read like blood. Not pink but red. Taking images of the lamp under artificial light results in the lava being rendered pink. With my Google Pixel (see below), but also with my Leica Q2, Sony a7R3A and Fuji X-E3. I have reduced exposure to avoid clipping the red channel. The lava itself is not very bright. How come that all sensors return pink instead of red?

Nielk Mike said:

Can someone help me with this? We have a nice lava lamp from the 70ies at home. The fluid is blue and the "lava" is read like blood. Not pink but red. Taking images of the lamp under artificial light results in the lava being rendered pink. With my Google Pixel (see below), but also with my Leica Q2, Sony a7R3A and Fuji X-E3. I have reduced exposure to avoid clipping the red channel. The lava itself is not very bright. How come that all sensors return pink instead of red?

Click to expand...

That is near-infrared light, which registers as strongly in all-three color channels in typical Bayer CFAs, so depending on your WB, it will render somewhere around magenta.

You would either need an addition near-IR-cut filter, or take a second shot with an IR-pass filter, and subtract it after optimally scaling it. to reduce this magenta.

Or, if you can change the incandescent bulb in the lamp for an LED with no IR, that could work, too, but the cooler bulb will stop the lava flow. A filter over the bulb that turned IR directly into heat might do the trick as well, without cooling the lamp .

--
Beware of correct answers to wrong questions.
John

John Sheehy said:

Nielk Mike said:

... Taking images of the lamp under artificial light results in the lava being rendered pink. .... I have reduced exposure to avoid clipping the red channel. The lava itself is not very bright. How come that all sensors return pink instead of red?

Click to expand...

That is near-infrared light, which registers as strongly in all-three color channels in typical Bayer CFAs, so depending on your WB, it will render somewhere around magenta.

Click to expand...

Mike, John is likely correct. However, mixed light sources is likely part of the problem.

However a simpler solution to get something "close" would be to shoot it with a known colour target such as a ColorChecker, grey card or just a piece of white paper. You also need to save the RAW file not just jpg.

Then use that to set the colour balance. This would likely not be "technically correct" but maybe close enough...

Richard

Thank you both. Will try.

John Sheehy said:

Nielk Mike said:

Can someone help me with this? We have a nice lava lamp from the 70ies at home. The fluid is blue and the "lava" is read like blood. Not pink but red. Taking images of the lamp under artificial light results in the lava being rendered pink. With my Google Pixel (see below), but also with my Leica Q2, Sony a7R3A and Fuji X-E3. I have reduced exposure to avoid clipping the red channel. The lava itself is not very bright. How come that all sensors return pink instead of red?

Click to expand...

That is near-infrared light, which registers as strongly in all-three color channels in typical Bayer CFAs, so depending on your WB, it will render somewhere around magenta.

You would either need an addition near-IR-cut filter, or take a second shot with an IR-pass filter, and subtract it after optimally scaling it. to reduce this magenta.

Or, if you can change the incandescent bulb in the lamp for an LED with no IR, that could work, too, but the cooler bulb will stop the lava flow. A filter over the bulb that turned IR directly into heat might do the trick as well, without cooling the lamp .

Click to expand...

With all respect, but I do not think that this is due to an infrared issue. Especially since the Google Pixel is a small sensor camera, which probably implies a infrared cutoff already at shorter red wavelength. Cameras have a spectral sensitivity which differs from the human eye: violation of Luther-Ives condition for faithful color reproduction across the board. The color processing is such that important, spectrally broad, colors such as sampled in a color checker card render close to perfect. The human observer is particularly picky about human skin tones. Colors with unusual spectral properties can get rendered far off the mark. -- I may get a such a rad lava lamp for curiosity. If you have got a photo-spectrometer like the Xrite iStudio 1 you may investigate the spectrum yourself.

In an earlier thread, I came across a case where an object was rendered with too much blue . With a shift from red to pink rendering for the lava-lamp, it is also a case of rendering a too strong blue component.

The spectral sensitivities for recent some Nikon camera models are grossly similar to other contemporary cameras. The spectral sensitivity goes to practically zero before the infrared. It would require a huge infrared component coming from the lava-lamp to overturn the color balance coming from the 420-680 nm spectral range.

Bernard Delley said:

John Sheehy said:

Nielk Mike said:

Can someone help me with this? We have a nice lava lamp from the 70ies at home. The fluid is blue and the "lava" is read like blood. Not pink but red.

Click to expand...

Click to expand...

Click to expand...

Bernard, it is not clear if you are responding to John, me, or both as you only quoted John... (IR; me, ColorChecker)

Bernard Delley said:

Bernard Delley said:

That is near-infrared light, which registers as strongly in all-three color channels in typical Bayer CFAs, so depending on your WB, it will render somewhere around magenta.

Click to expand...

With all respect, but I do not think that this is due to an infrared issue. ... Cameras have a spectral sensitivity which differs from the human eye: violation of Luther-Ives condition for faithful color reproduction across the board. The color processing is such that important, spectrally broad, colors such as sampled in a color checker card render close to perfect. ....Colors with unusual spectral properties can get rendered far off the mark.

Click to expand...

Yes, there are a number of cases where artificial illumination renders colours differently. For example sodium (orange) street lights make red firetrucks look black... not a good idea.

Then there are some flowers, esp blue, that are tricky to render...

Many of these are because of disjointed spectra (discharge lamps,) with gaps, or strong spikes, in the waveband.

Even our lowly incandescent photo lamps render differently to "daylight" (which itself varies..).

I guess there are two "solutions", figure out the problem in advance, and use filters or change the light source, or use a standard reference target and adjust in post-processing.

Then there are the cases where you actually don't want "technically correct" renderings, like sunsets... A Lava lamp my also be one of these as it is a "atmosphere shot".

Richard

Nielk Mike said:

Thank you both. Will try.

Click to expand...

You might try the lava lamp as sole source of light in the dark, and use white balance fixed at daylight. See what you get with this.

If you have the possibility, you could go on, and correct the color in post processing. Find out what hue shift is needed to show approximately the correct color on the screen.

Richard James said:

Bernard Delley said:

John Sheehy said:

Nielk Mike said:

Can someone help me with this? We have a nice lava lamp from the 70ies at home. The fluid is blue and the "lava" is read like blood. Not pink but red.

Click to expand...

Click to expand...

Click to expand...

Bernard, it is not clear if you are responding to John, me, or both as you only quoted John... (IR; me, ColorChecker)

Richard James said:

Richard James said:

That is near-infrared light, which registers as strongly in all-three color channels in typical Bayer CFAs, so depending on your WB, it will render somewhere around magenta.

Click to expand...

With all respect, but I do not think that this is due to an infrared issue. ... Cameras have a spectral sensitivity which differs from the human eye: violation of Luther-Ives condition for faithful color reproduction across the board. The color processing is such that important, spectrally broad, colors such as sampled in a color checker card render close to perfect. ....Colors with unusual spectral properties can get rendered far off the mark.

Click to expand...

Yes, there are a number of cases where artificial illumination renders colours differently. For example sodium (orange) street lights make red firetrucks look black... not a good idea.

Then there are some flowers, esp blue, that are tricky to render...

Many of these are because of disjointed spectra (discharge lamps,) with gaps, or strong spikes, in the waveband.

Even our lowly incandescent photo lamps render differently to "daylight" (which itself varies..).

I guess there are two "solutions", figure out the problem in advance, and use filters or change the light source, or use a standard reference target and adjust in post-processing.

Click to expand...

Some post-processing apps can adjust individual hues in an image without affecting others thereby turning pinks into reds, e.g. RawTherapee.

In some color literature, the hue of pink is actually red, So, one could select the lava and crank the sat. ..,.

xpatUSA said:

Richard James said:

I guess there are two "solutions", figure out the problem in advance, and use filters or change the light source, or use a standard reference target and adjust in post-processing.

Click to expand...

Some post-processing apps can adjust individual hues in an image without affecting others thereby turning pinks into reds, e.g. RawTherapee.

In some color literature, the hue of pink is actually red, So, one could select the lava and crank the sat. ..,.

Click to expand...

Yes, I do that often in PS.

But "colours" are (almost) never pure (single frequency or very narrow band), so there will always be some bleed... Whether that works well enough, or not , likely depends on the specific lighting.

Ultimately, having a known reference in the image is key.

Richard