Timeline for How much light and resolution is lost to color filter arrays?
Current License: CC BY-SA 4.0
28 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 16, 2020 at 21:20 | comment | added | Orbit | @MichaelC: The difference does not seem much in these plots. It does not seem likely that it makes a very big difference. | |
| Jun 16, 2020 at 12:27 | history | edited | Michael C | CC BY-SA 4.0 |
added 1349 characters in body
|
| Jun 16, 2020 at 12:13 | comment | added | Michael C | @Orbit Please see here and here | |
| Jun 11, 2020 at 18:15 | comment | added | Orbit | @MichaelC: Do you have have any source for this? I find that it is pretty evenly distributed around noon. Around sunset it goes more towards red, as we all know. But even if it is so, we still need blue and red to represent all colors. | |
| Jun 11, 2020 at 1:16 | comment | added | Michael C | @Orbit You're still assuming that "white light" has equal amounts of energy across the entire visible spectrum. It does not. Sunlight filtered by Earth's atmosphere, and the artificial light sources we create to mimic that light, has much more intensity in the mi- range (green) wavelengths than in the shorter and longer wavelengths on either side of the visible spectrum. Our retinas evolved to be most sensitive to the most energetic portion of "white light" and our cameras have sensors that mimic that. | |
| Jun 11, 2020 at 1:08 | history | edited | Michael C | CC BY-SA 4.0 |
edited body
|
| Jan 30, 2020 at 20:44 | comment | added | Orbit | @MichaelC: The Bayer filter does let around 50% pass at some wavelengths, but only 25% at others, that's why the total efficiency comes close to 1/3, although it is slightly higher than that because it has 2 green filters for every red or blue one. The 20% in last your sentence is quite accurate.That is the situation at 470 nm. There the monochrome efficiency is 68%, the filter efficiency is 32% and the total efficiency is 20%. That is a region where the eyes are quite sensitive for blue light, but red and green also start to become active. | |
| Jan 29, 2020 at 23:29 | comment | added | Michael C | So now saying some color sensors are "approaching 60%" somehow actually means we're claiming 70%? | |
| Jan 29, 2020 at 23:27 | comment | added | Michael C | No it doesn't. If the same sensor (from your own claims) when measured without a Bayer mask is 60% efficient, and it is 30% efficient with the Bayer mask, then the Bayer mask is only cutting 50% of the light hitting it. If it were cutting 2/3 of the light and what light gets through the Bayer mask was striking a sensor with 60% QE without a Bayer mask, then only 60% of the 33% that makes it through the Bayer mask would be measured, for a system QE of 20%. | |
| Jan 29, 2020 at 22:37 | comment | added | Orbit | @MichaelC: I never said that the 15% is normative, it was just an example. With a peak around 35% and also large regions where it is close to 15%, the average should be around 25%. That is nowhere near the 50 to 70% that you are claiming. So the conclusion is that a Bayer filter does cut out close to 2/3 of the light. | |
| Jan 29, 2020 at 0:19 | comment | added | Michael C | @Orbit You're spot selecting one of the most inefficient places over the entire visible spectrum for that sensor and then arguing that is normative. It is not. At 540µm, where our own eyes are most efficient, the difference between monochrome at 67% and green at 64% is negligible. Even assuming R and B are 0% (not the case), since half the sensels are filtered for green the entire sensor would still be 32% efficient at 540µm. See how that works? | |
| Jan 28, 2020 at 21:00 | comment | added | Orbit | @MichaelC: The monochrome efficiency is measured without bayer filter, therefore it is very high. With the filter it is dramatically lower. The total efficiency can be calculated for every wavelength as shown above. At 650 the monochrome efficiency is 45%, yet the total efficiency is only 15%, as shown above. | |
| Jan 25, 2020 at 18:27 | comment | added | Michael C | So the comparative color QE needs to be compared against that 60%, not an assumed 100%. | |
| Jan 25, 2020 at 18:23 | comment | added | Michael C | As you can see from the relatively old Sony IMX249 shown above, the peak monochrome efficiency is around 70% at 505µm and stays above 60% from 420-575µm Between about 400µm and somewhere just past 700µm, an average efficiency of 60% is maintained. That's almost the entire visible spectrum... | |
| Jan 25, 2020 at 18:18 | comment | added | Michael C | The reason there are two green pixels for each blue/red pair is that the human retina is similar. This is due to evolving in a world where the dominant light is much stronger in the "green" range of wavelengths closer to the center of the visible spectrum than in the extremes of the visible spectrum. When shooting under anything resembling sunlight/natural light (which most man-made artificial light sources are), our sensors are designed to be more sensitive to the ranges of wavelengths that are more plentiful. | |
| Jan 25, 2020 at 12:52 | comment | added | Orbit | Unfortunately the QE of 60% is a peak value, the average is much lower. If you look at a wavelength of 650 nm, the red channel passes 40%, the green channel 9% and the blue channel only 3%. That means that in total only 15% of the light with that wavelength is registered by the sensor, taking in to account that there are 2 green pixels for every blue or red pixel. | |
| Mar 2, 2017 at 18:44 | comment | added | Euri Pinhollow | @feetwet practically Bayer CFA does not cut resolution of monochrome objects given that they are not too saturated. 50% of QE cut is optimistic, it may be well less than that. | |
| Mar 2, 2017 at 14:27 | comment | added | feetwet | Did you not mean to say that a Bayer CFA in effect rejects 50% of visible light? If not, the question would be, "How much visible light is rejected by a typical CFA?" | |
| Mar 2, 2017 at 14:22 | comment | added | Michael C | @feetwet 50% only if one assumes the QE of a non-masked sensor would be 100%. | |
| Mar 2, 2017 at 14:20 | vote | accept | feetwet | ||
| Mar 2, 2017 at 14:20 | comment | added | feetwet | So TL;DR: In practice Bayer CFAs cut light by roughly 50% and resolution by about 30%. | |
| Mar 2, 2017 at 12:10 | comment | added | Euri Pinhollow | The quoted claim is disproved in this long comment: drive.google.com/file/d/0By7viOLQaKydMVUwTjVDT0JBSXM/… | |
| Mar 2, 2017 at 9:43 | comment | added | Michael C | I haven't quoted anything regarding monochrome QE. I referenced monochrome resolution limits. | |
| Mar 2, 2017 at 9:33 | comment | added | Euri Pinhollow | How was monochrome QE obtained then? | |
| Mar 2, 2017 at 9:19 | comment | added | Michael C | The QEs quoted are measured with the Bayer mask in place. | |
| Mar 2, 2017 at 8:00 | comment | added | Euri Pinhollow | The first graph is not insightful because, really, only a fraction of colour filters would have the QE of a channel. Divide blue, red graphs by 4 and green graph by 2, sum them and you will see that resulting QE graph (which is roughly equal to what you get if you use monochrome conversion in post) is magnitude smaller than monochrome QE. it does show that Bayer CFAs are not loosing much more than what they are expected to by design though. | |
| Mar 2, 2017 at 7:48 | history | edited | Michael C | CC BY-SA 3.0 |
added 350 characters in body
|
| Mar 2, 2017 at 7:28 | history | answered | Michael C | CC BY-SA 3.0 |
