Lenshacker Posted February 28, 2015 Share #161 Posted February 28, 2015 Advertisement (gone after registration) The IR sensitivity of the M9 is about 12 stops under the visible sensitivity. That is what I measured yesterday, and it agrees with the Data Sheet from Schott for the S8612 and the On Technology data sheets for their sensors. Fluorescent lighting has very little IR output in it, I just took a picture of one with the Full-Spectrum EP2. ISO 200, F3, 1/400th second with no filter, 1/5th second with the 88a filter. You can look up the spectrum for "fluorescent light spectrum cool white". If you are having color problems under Fluorescent lighting, it is not from IR contamination. If the lighting is mixed, where certain parts of the image are illuminated from sources with different balance- you need to either select one section of the image to correct, or do a piece-meal correction. LCT's original image- white balance problem, not IR contamination. For removing IR contamination from the image... Just do what you do with the M Monochrom to get Color images. But for the second image, put the 88a filter over the lens. Register the images and do a frame-to-frame subtract. Link to post Share on other sites More sharing options...
Advertisement Posted February 28, 2015 Posted February 28, 2015 Hi Lenshacker, Take a look here CCD vs CMOS: Can you tell which is which?{merged}. I'm sure you'll find what you were looking for!
CheshireCat Posted February 28, 2015 Share #162 Posted February 28, 2015 LCT's original image- white balance problem, not IR contamination. if that was really a WB problem, then why does the guy's collar look white In the original image ? Link to post Share on other sites More sharing options...
Lenshacker Posted February 28, 2015 Share #163 Posted February 28, 2015 Because you have mixed lighting with sources at different temperatures. The one that drives me crazy- window light+ fluorescent light, or Tungsten light + fluorescent light. Depending on the strength of each on a given point, just impossible to balance. The first couple times I used the M8, I packed the full-spectrum Coolpix 950 along with it. I wanted to see why certain objects had problems. Don't make me bore everyone again with how I called Kodak and had them make the first DCS200ir for me. It really was funny. "We spent years trying to get Silicon to look more like film, and you want us to undo it all". Link to post Share on other sites More sharing options...
AlanG Posted February 28, 2015 Share #164 Posted February 28, 2015 I see this as a mixed lighting issue. Maybe some sensors have slightly different responses but that is just as likely to be due to firmware or profiles. It is pretty hard to isolate. FWIW I had more trouble when shooting film and generally had to filter lights to one color as there was no way with film to adjust individual colors in post. Shooting digitally and using software such as DXO gives control over each color and allows me to work in mixed lighting easily. You can't expect color balance or a camera profile to handle all situations. You really have to learn to understand what is "wrong" with an image on a deeper level than just color balance before you can choose the tools needed to adjust it. For instance I saw here that some thought the issue was a green face and compensated with magenta. But from experience I knew it was localized yellow that was causing the main problem. Once you know that, the fix is easy. However this level of understanding light, color and software is not trivial and once was only mastered by experts and pre-press speciallists. Link to post Share on other sites More sharing options...
CheshireCat Posted February 28, 2015 Share #165 Posted February 28, 2015 The explanation of multiple light sources is not convincing. Why would the collar have a different "mix" than the face ? I found the following spectral responses for a few Canon cameras: EOS40D / EOS50D comparison Comparing them to the one of the M9, there is one notable difference in the M9 graph: the red graph stays well above 0 in the blue domain, and has a notable spike on blue wavelenght of about 380. Seems to me that a blue wavelenght of 380 is detected as blue=0.1 and red=0.07 (i.e. 100% blue and 70% red after normalization) Link to post Share on other sites More sharing options...
AlanG Posted February 28, 2015 Share #166 Posted February 28, 2015 I think this is nice, but I think the gentleman has some more red detailing in his face in reality. Could be but I never saw him in person and just went for a pleasant balanced look. I could get any look you want. I simply worked from the green image of the two posted here. A raw file might have worked better but is irrelevant for the purpose of illustration of the need to identify and isolate a color issue. Link to post Share on other sites More sharing options...
AlanG Posted February 28, 2015 Share #167 Posted February 28, 2015 Advertisement (gone after registration) The explanation of multiple light sources is not convincing. Why would the collar have a different "mix" than the face ? : Well then why would my adjustments work? The collar is white so it is hard to see uneven colors in it. Fabric and skin reflect colors differently. Link to post Share on other sites More sharing options...
lct Posted February 28, 2015 Share #168 Posted February 28, 2015 I only see oversaturated reds in the raw file below but i may be wrong and it is an early pic (10 sept 2009) anyway so perhaps the red cast has been removed in a later firmware updates i don't know. http://tinyurl.com/md9cxn7 (18MB file) Link to post Share on other sites More sharing options...
Lenshacker Posted February 28, 2015 Share #169 Posted February 28, 2015 The explanation of multiple light sources is not convincing. Why would the collar have a different "mix" than the face ? I found the following spectral responses for a few Canon cameras: EOS40D / EOS50D comparison Comparing them to the one of the M9, there is one notable difference in the M9 graph: the red graph stays well above 0 in the blue domain, and has a notable spike on blue wavelenght of about 380. Seems to me that a blue wavelenght of 380 is detected as blue=0.1 and red=0.07 (i.e. 100% blue and 70% red after normalization) The Canon spectral plots stop at 400nm. Either they did not plot into the UV to see what would happen, or they incorporate a UV filter into the optical path. Do some tests: put out some white poster board with a Tunsgten light on one end and a Fluorescent light on the other. Put some objects along the way. Should be an interesting test to see what comes out. Color Dye "goes all over the place" in IR and UV. Does not surprise me that the Red passes UV as the Peak is farthest from it. The spectral response of the KAF-18500 shows the importance of using a UV filter. I went through a closet full of camera stuff looking for the 87c filter. Didn't find it- found the RCA CKC-020 video camera bought over 30 years ago. One of the very first CCD cameras on the market, bought it to shoot video at Air Shows. Figured the solid state sensor would not burn out as a Vidicon tube does. Link to post Share on other sites More sharing options...
AlanG Posted February 28, 2015 Share #170 Posted February 28, 2015 A glance at the man's hair should tip off anyone that there is mixed lighting. So step one is identifying the problem. This is pretty basic and I only felt the need to post after being frustrated by all of the teeth grinding here. Assuming the sensors and overall systems record adequate separation images, then it is up to software how to get them to depict the scene the way you want it. In this case the "good" result differs from the reality. I am sure some systems give more capability than others but you can't isolate those differences with casual examples. The camera could not know what you want nor is it capable of doing that on its own. Color photography has always had the challenge of the difference between perception and reality. And until you prove that a given file cannot be adjusted to get the results you want, why would you expect another camera to do better? Link to post Share on other sites More sharing options...
jaapv Posted February 28, 2015 Share #171 Posted February 28, 2015 No way. If that is actually IR-bleed there is nothing to do about it, and any solution will be a compromise. The explanation is simple. Imagine you have two pixels A and B, both with the same value of red=100. But the value of A was given by 90 visible-red + 10 IR-bleed. And the value of B was given by 50 visible-red + 50 IR-bleed. Good luck with your LAB corrections. You cannot restore information that has been lost. Granted, you (as a human) may guess the IR bleeding for the different pixel areas and apply different corrections with several masks until the result seems natural. But using the same time, I could rather paint oil on canvas for a much nicer result In LAB you can even create colurs that do not exist...I agree that it takes a lot of time, though. Link to post Share on other sites More sharing options...
jaapv Posted February 28, 2015 Share #172 Posted February 28, 2015 I see this as a mixed lighting issue. Maybe some sensors have slightly different responses but that is just as likely to be due to firmware or profiles. It is pretty hard to isolate. FWIW I had more trouble when shooting film and generally had to filter lights to one color as there was no way with film to adjust individual colors in post. Shooting digitally and using software such as DXO gives control over each color and allows me to work in mixed lighting easily. You can't expect color balance or a camera profile to handle all situations. You really have to learn to understand what is "wrong" with an image on a deeper level than just color balance before you can choose the tools needed to adjust it. For instance I saw here that some thought the issue was a green face and compensated with magenta. But from experience I knew it was localized yellow that was causing the main problem. Once you know that, the fix is easy. However this level of understanding light, color and software is not trivial and once was only mastered by experts and pre-press specialists. Simple measuring helps a lot in identifying the problem. The main problem is twofold. There is probably more than one cast and human skin it a subject that produces a considerable amount of metamerism, due to its layered structure. Link to post Share on other sites More sharing options...
Lenshacker Posted February 28, 2015 Share #173 Posted February 28, 2015 Full-Spectrum camera with a green-notch filter (Magenta in color, absorbs Green band). DSCN0178 Me and Color Balance... Which is why I bought an M Monochrom from David. Link to post Share on other sites More sharing options...
AlanG Posted February 28, 2015 Share #174 Posted February 28, 2015 I found this link very illuminating: 'You Are Not So Smart': Why We Can't Tell Good Wine From Bad - The Atlantic I think it applies to a lot of things such as cameras and lenses sold with brand identification and also high status items, film vs. digital quality, CCD vs CMOS, etc. Link to post Share on other sites More sharing options...
TonyField Posted March 1, 2015 Share #175 Posted March 1, 2015 Full-Spectrum camera with a green-notch filter (Magenta in color, absorbs Green band).Me and Color Balance... Which is why I bought an M Monochrom from David. I think this implies that, in spite of a strong green cut filter, the image residual green is, for the most part (as determined by the quality of the filter), IR contamination of the green pixels. Link to post Share on other sites More sharing options...
jaapv Posted March 1, 2015 Share #176 Posted March 1, 2015 I think a better way would be to compare the exposure time needed without filter and with a 093 filter which gives a sharp cutoff of all visible light. Link to post Share on other sites More sharing options...
Lenshacker Posted March 1, 2015 Share #177 Posted March 1, 2015 I think a better way would be to compare the exposure time needed without filter and with a 093 filter which gives a sharp cutoff of all visible light. Do you have the 093 filter? If so- can you try the test with the M240? The B&W 093 is a Wratten 87c, block everything under 790nm. I have one other place to search for it! I use the 88a with the full-spectrum camera, the 87c is much better for unmodified cameras. I was using Digital Infrared and full-spectrum cameras 15 years before I bought my first visible-only digital camera, a Kodak DC50. The S8612 cutoff has already ramped up ( the S8612 data sheet indicates 2.9E-3 transmission at 720nm) where the 88a begins, at 720nm. The 87 is 740nm, well into the S8612 cut-off. Someone made the transmission numbers for Wratten filters into a PDF: http://www.karmalimbo.com/aro/pics/filters/transmision%20of%20wratten%20filters.pdf I have the old data books from the 1960s. Link to post Share on other sites More sharing options...
Lenshacker Posted March 1, 2015 Share #178 Posted March 1, 2015 I think this implies that, in spite of a strong green cut filter, the image residual green is, for the most part (as determined by the quality of the filter), IR contamination of the green pixels. Yes- I was going for an "Ektachrome E3" look. Grass came out Red, and sky came out blue. Of course, "IR" is not considered contamination in a full-spectrum camera. More like the IR cutoff filter is considered a nuisance. Link to post Share on other sites More sharing options...
Lenshacker Posted March 1, 2015 Share #179 Posted March 1, 2015 The human eye cannot see out to 720nm, stops just past 700nm. The 88a filter cuts out all of the wavelengths that you can actually see. It's probably the best filter for testing if the amounts of IR that the camera picks up that can cause colors to differ from what we actually see. Link to post Share on other sites More sharing options...
jaapv Posted March 1, 2015 Share #180 Posted March 1, 2015 Do you have the 093 filter? If so- can you try the test with the M240? The B&W 093 is a Wratten 87c, block everything under 790nm. I have one other place to search for it! I use the 88a with the full-spectrum camera, the 87c is much better for unmodified cameras. I was using Digital Infrared and full-spectrum cameras 15 years before I bought my first visible-only digital camera, a Kodak DC50. The S8612 cutoff has already ramped up ( the S8612 data sheet indicates 2.9E-3 transmission at 720nm) where the 88a begins, at 720nm. The 87 is 740nm, well into the S8612 cut-off. Someone made the transmission numbers for Wratten filters into a PDF: http://www.karmalimbo.com/aro/pics/filters/transmision%20of%20wratten%20filters.pdf I have the old data books from the 1960s. When I have time, my M9 back from Wetzlar for comparison and some IR in the light. Link to post Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.