I've written this to the best of my knowledge but I'm sure there are some inaccuracies in it; helpful suggestions for improvements (whether Scandinavian in origin or not) are welcome!
What a raw image really looks like
It is important to remember that the sensor sites ("pixels") can only distinguish tonality and not frequency of light. Hit the sensor with a certain amount of energy and it will read out a certain discrete value. It doesn't matter if that energy is a few photons of high intensity (blue light) or a whole bunch of lower intensity (red light)--energy is energy.
Here is a bayer filter for a 2k (46x46 pixel) sensor:
![Posted Image]()
And here's the image we're trying to capture with it (after the image passed through our spotless low pass (anti aliassing) filter:
![Posted Image]()
The bayer filter reduces the light for each pixel to either the red, green or blue band. After passing through the filter the image looks like this:
![Posted Image]()
But remember, the sensor really doesn't care about colors, it only looks at energy levels in a single channel, like this:
![Posted Image]()
At this point it's worth noticing that we only get a single readout per pixel, like this:
![Posted Image]()
This (above) is what the raw file contains. Each of the RGB "pixels" contains a single gray-scale value. It's incorrect to see a mid-gray pixel value as RGB #777777 because there is no RGB, just grey values. At this point the sensor contains as much color information as Kodak T-MAX film, that is to say, none (yes, TMAX is more purplish than grey when developed but it doesn't contain information about the original color). Well, we do have some information about the colors since we know for each pixel what the corresponding bayer filter (R, G or B) is.
Now we run this data through our EXPEED, BIONZ or whatever fancy acronym processor in the camera, or through our beloved raw converter on the computer, and after a bit of witchcraft we get this:
![Posted Image]()
Obviously, this image contains much more information than the previous one - 3 times as much, to be precise. I will leave it to the reader to ponder over this (hint: "Claude Shannon") and the relevancy of pixel peeping at that point.
Comment on this article
Editors comment:
Bart Willems is a New Jersey based amateur photographer who likes to share his knowledge, you might find his Blog worth a visit.
What a raw image really looks like
It is important to remember that the sensor sites ("pixels") can only distinguish tonality and not frequency of light. Hit the sensor with a certain amount of energy and it will read out a certain discrete value. It doesn't matter if that energy is a few photons of high intensity (blue light) or a whole bunch of lower intensity (red light)--energy is energy.
Here is a bayer filter for a 2k (46x46 pixel) sensor:
And here's the image we're trying to capture with it (after the image passed through our spotless low pass (anti aliassing) filter:
The bayer filter reduces the light for each pixel to either the red, green or blue band. After passing through the filter the image looks like this:
But remember, the sensor really doesn't care about colors, it only looks at energy levels in a single channel, like this:
At this point it's worth noticing that we only get a single readout per pixel, like this:
This (above) is what the raw file contains. Each of the RGB "pixels" contains a single gray-scale value. It's incorrect to see a mid-gray pixel value as RGB #777777 because there is no RGB, just grey values. At this point the sensor contains as much color information as Kodak T-MAX film, that is to say, none (yes, TMAX is more purplish than grey when developed but it doesn't contain information about the original color). Well, we do have some information about the colors since we know for each pixel what the corresponding bayer filter (R, G or B) is.
Now we run this data through our EXPEED, BIONZ or whatever fancy acronym processor in the camera, or through our beloved raw converter on the computer, and after a bit of witchcraft we get this:
Obviously, this image contains much more information than the previous one - 3 times as much, to be precise. I will leave it to the reader to ponder over this (hint: "Claude Shannon") and the relevancy of pixel peeping at that point.
Comment on this article
Editors comment:
Bart Willems is a New Jersey based amateur photographer who likes to share his knowledge, you might find his Blog worth a visit.