Photography Stack Exchange is a question and answer site for professional, enthusiast and amateur photographers. It only takes a minute to sign up.
Sign up to join this community
I recently photographed an event that took place while the sun was setting. The quality of colors seem to deteriorate as it gets darker. The scenes were well-exposed as the ISO changed from ~1200 to ~6400.
Is the color quality change simply due to an increase in ISO, and if so, why?
Does color shift in a particular direction when light intensity drops?
Does red/blue color quality deteriorate faster due to bayer filter implementations?
High ISO can be a factor (see Is there a technique to increase saturation in high ISO?), but I suspect that another effect you are seeing is the decrease of natural-full spectrum light relative to the fluorescent light source — see How does light quality vary between fluorescent (CFL) and incandescent? (with sunlight being roughly the same as incandescent in this regard).
Basically, the sunlit photos include all the colors in the scene, even though there may be some white balance shifts. Later, as the photos are primarily lit by fluorescent tubes, the colors that don't respond to wavelengths in that light source don't show up, causing the "deterioration" you describe.
I think this depends on what you mean by "quality of colors". Fundamentally, if you have a block of pure colour and photograph it at ISO 100, you'll get something which is just about pure colour out of the camera - or at least, every red/green/blue photosite on the sensor will record the same value as every other red/green/blue photosite on the sensor; how that's actually translated to an RGB value is question of a few other things, principally white balance. While there obviously be a small amount of noise at ISO 100, it's going to be pretty small on a modern camera.
On the other hand, if you photograph the same block of pure colour at ISO 6400, there's going to be a significant amount of noise from the sensor, so different photosites will record different values. This means that your block of pure colour will no longer be pure colour, but will have random noise in it which makes it look different. If the magnitude of the noise is similar to that of the signal in the first place, you're not going to get something which looks like a pure colour any more.
Sensor responses are relatively linear, the issue is in processing the image and in handling the noise. As the signal decreases and ISO increases, a greater electrical charge is applied to the sensor to make it more sensitive to light. Think of it as a bayer pattern of buckets. In average light, there is enough light to fill the buckets without much help fro ISO, but in low light/high ISO conditions the electric charge is increased and less light is needed to create a full bucket. This noisy light contains color noise that most manufacturers try and process away, but some lighting situations may cause that noise to be more pronounced.
So the short answer to all of your questions is yes, but the direction of the shift in error depends on the spectral sensitivities of the sensor and the light coming through the lens. The Red/ blue (mostly blue) does deteriorate more quickly than the green because those filter colors have higher additivity failure. The blue channel is the noisiest for that same reason.
It seems like an evidence but shoot in raw format in this case because:
color temperature change and jpeg need to preset the temperature reference so once shoot and write in the file, the original information is lost, translate to the new reference. It is hard to rechange to another reference without loosing more info. Raw files are the information directly output by the sensor, with few interpretation (protocol, file format/info are added info, ...) and color temperature is just an info telling how to deal with the content but this can be changed at creation of the jpg (or other format) via the tools you use to extract picture.
Raw format allow (and use) a bigger scale to store elementary color information (more bit) that allow to have more subtility that are very interresting at the edge of the scale. The information is not available in jpeg (a average value making big step in perception in lower light that eye could easily see). Post processing is needed to use this information in raw but with the possibility to rescale with your own average taht will better suite, from your perception pint of view, to the picture.
Why is it nearly mandatory in sunset: For the hight ISO, the modification of ISO is in fact an amplification of the signal of the cell from the sensor, not like the old roll film that change the nature of the sensor itself. This means that noise (electronic concept) is also amplyfied and other specific problem appear (especialy for long exposure). This will modify color and light perceived by the sensor (the output of the sensor, sensor is not changing the light ray). This will create a difference with real color/light and the write info into the file.
High ISO are interessting to still take a picture when light is low but the cost is the noise in color and intensity. The noise is a bit hazardous and 2 shoot in same condition will give you 2 different picture where low ISO (low mean the nearest value to the real sensitivity of the sensor, not the lowest available on the camera) will give (with enough light) 2 nearly identique picture.
You also need to know that, as you know, lower ISO will also say longer exposure at same aparture, have another side effect on sensor. The longer the sensor is "open", the more heat is generated and heat have a similar effect on the output than amplification, it create noise.
So, you need to find the optimum between ISO and exposure time to reduce the perturbation.
Last important point is the way modern sensor perceive the color. Appart some special sensor like the foveon of sigma, all sensor can only see light intensity, not color light intensity nor color itself. A specific matric of colored micro lens at put in front off sensor (bayer matrix) and color is interpreted for this information. (schema of a bayer matrix) Because each elemetary color is not see the same way depending intensity, the interpreted value is giving the bad info (ex: if on 2 cell sensor see 1 red at 75% and 1 green at 75% that give a green olive, reducing the light could give a red at 30% and a green at 15% giving dark brown and not a dark green olive)
