 |
 | Home | Contact Us | FAQ | Search & Site Map | Link to Us |
 | Sign In | Join | Other 45 Sites in Network |
| PhotoKB HomeDiscussion Groups Digital Photography Digital PhotoDSLR CamerasZLR CamerasPoint & Shoot CamerasFilm Photography 35 mmLarge FormatMedium formatDarkroomFilm and LabsOther EquipmentPhoto Technique Nature PhotographyPeople PhotographyTechnique GeneralGeneral Photo Topics General TopicsAustralian PhotographyUK PhotographyDirectoryPhoto Clubs |
Photo Forum / Digital Photography / DSLR Cameras / October 2006Higher dynamic range. Mfgs (except Fuji) ignoring it. Why?
Pixels are still the same, square. Fuji has tried to deal with the issue. But Canon, Nikon, etc, don't seem to care as much about it as they do pixel counts. Why is this the case? It would be nice to be able to shoot a sunlit scene with shadowed areas knowing it won't take heavy manipulation to make the image as good as it can be. Or, upon bringing up the illumination in the shadowed areas, not having them noisy or devoid of good colour and tonality. P.S. Those HDR compilation shots people do just look unnatural. > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 4 lines] > noisy or devoid of good colour and tonality. > P.S. Those HDR compilation shots people do just look unnatural. If you are shooting with a DSLR at ISO 100 and shooting raw this is rarely a problem. Having said that I am a bit disappointed that they don't use more then 12 bit A/D converters. Scott > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 4 lines] > noisy or devoid of good colour and tonality. > P.S. Those HDR compilation shots people do just look unnatural. It's germane to note that Fuji seem stuck at 6 Mpix in the extended dynamic range sensors. For them to get to a decent 8 or a very decent 10 Mpix would require 16 - 20 Msensors ... tough act even if the high dyn sensors are smaller compared to the normal sensor. On the other hand many, many photographers do wonderful work with the current middle + 2 stops... eg: same as slide film. Plus the benefit of a stop or two more shaddow detail than slide film. Fuji's approach is great, but eliminating one problem has put them in last spot in the Mpix race ... which seems to be petering out at around 10 - 12 Mpix for the high-end prosumer bodies. I want to experiment with Sony's HDR mode. I've seen some examples posted, but by a relatively inexperienced photographer. The A100 just is not the right body for me. Hopefully they'll do a Maxxum 7D or Maxxum 9 class machine. Then I'll buy one.  Signature-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm -- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm -- [SI] gallery & rulz: http://www.pbase.com/shootin -- e-meil: Remove FreeLunch. > > Pixels are still the same, square. Fuji has tried to deal with the > > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 22 lines] > is not the right body for me. Hopefully they'll do a Maxxum 7D or > Maxxum 9 class machine. Then I'll buy one. Sony (at a show) told me more DSLRs, higher end are coming. > Sony (at a show) told me more DSLRs, higher end are coming. Glad to hear it. I hope they do right. They've been short to date. Signature-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm -- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm -- [SI] gallery & rulz: http://www.pbase.com/shootin -- e-meil: Remove FreeLunch. I don't think manufacturers are ignoring the dynamic range limits of digital sensors so much as their have not been any breakthroughs in materials and design techniques. It will happen long before there is cure for cancer, rest assured. It would appear that the moderate increase in dynamic range that the Fuji sensors provide is offset by other issues with the sensor and with Fuji dSLRs. Certainly Fuji dSLRs have only a limited market share which would suggest that for practical purposes photographers do not see great value to Fuji's claim of greater dynamic range. While some printed materials state that current digital imaging sensors have only one tenth of an f-stop latitude for over-exposure many photographers claim up to 11 f-stops (they claim it anyway) in their finished images. Anyone someone adept at shooting raw realizes that using the transparency film formula of exposing for the highlights allows detail from moderately underexposed areas to be brought up with acceptable noise levels. Whether this is "excessive" manipulation is a matter of personal attitude. Outside of controlled studio conditions all images require some manipulation to achieve an optimal print. If you do comparison shooting with ISO 200 color negative film and current dSLRs (same subject/lighting/lens and high quality scans with a dedicated film scanner) you may find, as I was surprised to find, that the often ballyhooed claim of massive dynamic range for color negative film just isn't, like many things in life, what it is supposed to be. It wasn't until I did that exercise that I truly realized that film had reached the end of its run as the imaging medium of choice. > I don't think manufacturers are ignoring the dynamic range limits of digital > sensors so much as their have not been any breakthroughs in materials and [quoted text clipped - 5 lines] > suggest that for practical purposes photographers do not see great value to > Fuji's claim of greater dynamic range. It's not a claim it's a ~2 stop (mid + 4) reality that comes at a price: reduced ability to increase pixel count. Increasing to 8 Mpix would be near meaningless in the resolution sense and 10 Mpix would require 20 M sensors (half "normal" / half highlight). > While some printed materials state that current digital imaging sensors have > only one tenth of an f-stop latitude for over-exposure many photographers [quoted text clipped - 5 lines] > of controlled studio conditions all images require some manipulation to > achieve an optimal print. This "manipulation" is a technical step akin to using the zone system to the limit of the technology in B&W. 11 stops in a finished image? Really? And where are they showing/printing this? > If you do comparison shooting with ISO 200 color negative film and current > dSLRs (same subject/lighting/lens and high quality scans with a dedicated [quoted text clipped - 3 lines] > I did that exercise that I truly realized that film had reached the end of > its run as the imaging medium of choice. That's somewhat true of neg-color film, but the common comparison (valid or not) is v. color slide film where, for the highlight end, digital and slide are remarkably similar, but digital digs down deeper. Neg color film still has at least one more stop to the higlights to offer, but in exposing that high the shaddows tend to be muddy looking. Cheers, Alan. Signature-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm -- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm -- [SI] gallery & rulz: http://www.pbase.com/shootin -- e-meil: Remove FreeLunch. > It's not a claim it's a ~2 stop (mid + 4) reality that comes at a > price: reduced ability to increase pixel count. Increasing to 8 Mpix > would be near meaningless in the resolution sense and 10 Mpix would > require 20 M sensors (half "normal" / half highlight). As I understand it, they don't have a double amount of sensors, instead more like one third more: A "normal" Bayer sensor has a red, green and blue pixel. Fuji adds an "S" pixel, making groups of four from what used to be groups of three. > If you do comparison shooting with ISO 200 color negative film and current > dSLRs (same subject/lighting/lens and high quality scans with a dedicated > film scanner) you may find, as I was surprised to find, that the often > ballyhooed claim of massive dynamic range for color negative film just > isn't, like many things in life, what it is supposed to be. It wasn't until > I did that exercise that I truly realized that film had reached the end of > its run as the imaging medium of choice. This is what a number of people are finding, that in fact a DSLR has more not less range then color print film. Roger Clark did a very good set of exparments dealing with this. Whereas a number of people has questioned Roger's experments none has done their own that show a different result. And here is a very good article that among other things talks about the pain of trying to get a lot of range out of color film. http://www.luminous-landscape.com/columns/1Ds-4x5.shtml Scott > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 4 lines] > noisy or devoid of good colour and tonality. > P.S. Those HDR compilation shots people do just look unnatural. If you shoot RAW you get 7 stops easily; if you push it a bit and use NR software you can get 9 stops. This is significantly more dynamic range than slide film, and a good stop or so more than B&W or negative film. That said, HDR shots do not *have* to look wonky. In fact, it's really an element of people not knowing how to properly adjust contrast (curves) in the highlight and shadow areas of an image -- they need different contrast curves. The unfortunate thing is that doing this with Photoshop is tricky, and requires some knowledge and practice. What you see with the HDR software is efforts for "the common person" and the end results are not good. I have seen work by people who are good with Photoshop (Marc Muench) and the results were fantastic, and looked nothing like HDR image. He just takes one shot, runs it through ACR twice (once exposed for shadow, once for highlights), combines, and works with the curves. So this is with one RAW shot -- the dynamic range is there -- with a 1Ds Mark II. > > Pixels are still the same, square. Fuji has tried to deal with the > > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 14 lines] > (curves) in the highlight and shadow areas of an image -- they need > different contrast curves. Actually, it's not. When you expose for a scene to capture shadow detail in an HDR shot, the bright areas bloom, and you can't stop it. Edges become diffuse looking and that gives the shot the weirdness it displays. There may be a way to deal with it in software. The idea that colour negative film has more latitude has to do with a latitude "cap" it imposes that keeps brighter areas from burning out upon overexposure (they just hit a density maximum but don't seem to go blank) as they do with digital shots. For shadow detail, (such as it is with it's increased noise and flatness) digital is superior. > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 4 lines] > noisy or devoid of good colour and tonality. > P.S. Those HDR compilation shots people do just look unnatural. Grad filters work & so does using flash. Get use to it. SignatureReality-Is finding that perfect picture and never looking back. www.gregblankphoto.com > Grad filters work & so does using flash. Get use to it. Grad filters work for simple scenes (horizons being the easiest examples), but not for deep shaddow scenes with light filtering through where the depth and volume of coverage is too great for flash to fill effectively. Signature-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm -- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm -- [SI] gallery & rulz: http://www.pbase.com/shootin -- e-meil: Remove FreeLunch. > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about > it as they do pixel counts. Why is this the case? Because DR mostly benefits shadows (because highlights are photon shot noise limited), and not all properly exposed ('to the right') shots require it, and because most prints are too small to benefit a lot. Also, (Mega)Pixel counts sell better. > It would be nice to be able to shoot a sunlit scene with shadowed > areas knowing it won't take heavy manipulation to make the image > as good as it can be. Or, upon bringing up the illumination in the > shadowed areas, not having them noisy or devoid of good colour > and tonality. The Dynamic range involved in such scenes, possibly exceeds all potential capture capabilities required and also exceeds the potential of the output media (print/display). > P.S. Those HDR compilation shots people do just look unnatural. Which has a lot to do with the people doing it (and the software choices they make). SignatureBart >> Pixels are still the same, square. Fuji has tried to deal with the >> issue. But Canon, Nikon, etc, don't seem to care as much about [quoted text clipped - 14 lines] >potential capture capabilities required and also exceeds the potential >of the output media (print/display). True, but "better" is better than nothing. Being about to have one more or even half a stop more on the top end would be good. Sony's new system in their A100 may have some benefit, but not much has been written about it yet. >> P.S. Those HDR compilation shots people do just look unnatural. > >Which has a lot to do with the people doing it (and the software >choices they make). But like dust cleaning, wouldn't it be so much better if this convoluted nonsense to obtain images wasn't required and the DR capability was "in-camera" to begin with? A solution to HDR could be a semi transparent mirror, reflecting 1/8 or 1/16 to a second sensor. Both sensors can be the same, this will improve the dynamic range with about 3 ro 4 stops. Offcourse the sensors have to be aligned critically. I would not know about the practical difficulties of this scheme, but 3 sensors are used for video. 2 would be possible for digital still camera's. (Actually the mirror should reflect 1/9 and pass through 8/9 or 1/17 and pass through 16/17 to get exactly 3 or 4 stops of difference between the two sensors). ben > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 4 lines] > noisy or devoid of good colour and tonality. > P.S. Those HDR compilation shots people do just look unnatural. > A solution to HDR could be a semi transparent mirror, > reflecting 1/8 or 1/16 to a second sensor. Both sensors [quoted text clipped - 9 lines] > 1/17 and pass through 16/17 to get exactly 3 or 4 stops > of difference between the two sensors). Ben, I think that is a very neat idea. Alignment could be done in software pretty effectively. For most cases I think large pixel cameras with 16-bit A/D converters are the way to go, but for special applications, your idea is pretty cool. Why stop at 4 stops? Make it a 6 stop difference. A large pixel camera with an 80,000 electron well and 4 electron read noise (14.3 stops of dynamic range could be extended to more than 20 stops). Of course some people will still complain that film is better. ;-) Roger > Pixels are still the same, square. Fuji has tried to deal with the > issue. But Canon, Nikon, etc, don't seem to care as much about it as [quoted text clipped - 4 lines] > noisy or devoid of good colour and tonality. > P.S. Those HDR compilation shots people do just look unnatural. Dynamic range is solely about the number of photons collected compared to the readout noise in the system. Example: the Canon 1D Mark II has a full well of about 80,000 electrons, with a read noise of only 3.9 electrons, for a dynamic range of 80,000 / 3.9 = 20500 or 14.3 stops (actual camera is limited by the 12-bit A/D). If read noise were reduced to 1 electron, the dynamic range would be an astonishing 80,000. Canon has put their effort in lowering the noise, which has other benefits too. Most other camera manufacturers have not achieved such low noise levels. If Canon split the sensor like Fuji did to give a smaller "less sensitive" pixel beside each larger pixel, they would not get as many photons in the larger pixel, reducing the dynamic range. It is not clear if such a strategy works very well. It should in theory, but calibration of the different pixels is difficult, and the benefits are small compared to the problems introduced. More on dynamic range: http://www.clarkvision.com/imagedetail/index.html#sensor_analysis Roger > Dynamic range is solely about the number of photons collected > compared to the readout noise in the system. > Example: the Canon 1D Mark II has a full well of about 80,000 > electrons, with a read noise of only 3.9 electrons, for a dynamic > range of 80,000 / 3.9 = 20500 or 14.3 stops (actual camera > is limited by the 12-bit A/D). Real cameras have lenses as well as sensors, and lenses have flare. If an area of the sensor somewhere is fully saturated at 80k electrons, is it really possible for another area of a sensor to be totally dark? How much dynamic range can be created in a single image, anyway? Andrew. >>Dynamic range is solely about the number of photons collected >>compared to the readout noise in the system. [quoted text clipped - 10 lines] > > Andrew. Yes it is possible. Sunrises/sunsets are good examples. The sun gets completely blown and saturates the sensor, but other parts of the image are extremely dark. For example, see: Digital Camera Raw Converter Shadow Detail and Image Editor Limitations: Factors in Getting Shadow Detail in Images http://www.clarkvision.com/imagedetail/raw.converter.shadow.detail Roger >> "Roger N. Clark (change username to rnclark)" <username@qwest.net> >> wrote: [quoted text clipped - 23 lines] > > Roger Here's a link to my own results after conversion using the same test image that you sent to me some time back using CS2: http://www.pbase.com/markuson/image/54914811/original Mark SignatureImages (Plus Snaps & Grabs) by Mark² at: www.pbase.com/markuson >> Digital Camera Raw Converter Shadow Detail and Image >> Editor Limitations: Factors in Getting Shadow Detail in Images [quoted text clipped - 5 lines] > > Mark Mark, I do remember this. If I remember correctly, you got that by tuning the converter for the shadows, and as a result compromising the highlights, correct? I was trying to get one overall best image with one conversion. After all the raw data information is there and it seems bizarre that going from 12-bit raw to 16-bit integers one would lose information. If I'm interpreting your example correctly, it shows that to get the full dynamic range with converters like the CS2 one, we must do two conversions, one for highlights, one for shadows, then blend the two. And that is what some people say they do. Roger >>> Digital Camera Raw Converter Shadow Detail and Image >>> Editor Limitations: Factors in Getting Shadow Detail in Images [quoted text clipped - 20 lines] > > Roger My example (the top portion) was the result of a single CS2 conversion only. SignatureImages (Plus Snaps & Grabs) by Mark² at: www.pbase.com/markuson >>>>Digital Camera Raw Converter Shadow Detail and Image >>>>Editor Limitations: Factors in Getting Shadow Detail in Images [quoted text clipped - 22 lines] > > My example (the top portion) was the result of a single CS2 conversion only. Mark, So can you tell me the settings you used, and/or show the whole frame? Thanks, Roger >>>>> Digital Camera Raw Converter Shadow Detail and Image >>>>> Editor Limitations: Factors in Getting Shadow Detail in Images [quoted text clipped - 26 lines] > So can you tell me the settings you used, and/or show > the whole frame? I was trying to remember that... It's no longer fresh in my mind. I'll do some hunting on my computer and see if I can come up with the files, but I suspect that once I posted my result...I didn't keep everything. I'll check... Mark SignatureImages (Plus Snaps & Grabs) by Mark² at: www.pbase.com/markuson >>>Dynamic range is solely about the number of photons collected >>>compared to the readout noise in the system. [quoted text clipped - 8 lines] >> totally dark? How much dynamic range can be created in a single >> image, anyway? > Yes it is possible. Sunrises/sunsets are good examples. > The sun gets completely blown and saturates the > sensor, but other parts of the image are extremely dark. > For example, see: > Digital Camera Raw Converter Shadow Detail and Image > Editor Limitations: Factors in Getting Shadow Detail in Images > http://www.clarkvision.com/imagedetail/raw.converter.shadow.detail Well, there are nearly 10 stops here, which might be getting close to the limit for real camera lenses. Multi-coating gets down to < 0.5% reflectance per surface, and there may be ten or more surfaces. The insides of a camera are all painted black, but there's still a significant reflectance; a few percent. Do you think, perhaps, that lens flare might limit true dynamic range before the limits of sensors are reached? [I presume that some astronomers know exactly what the practical limit of flare is for real lenses; I don't.] Andrew. .com/imagedetail/raw.converter.shadow.detail > Well, there are nearly 10 stops here, which might be getting close to > the limit for real camera lenses. Multi-coating gets down to < 0.5% [quoted text clipped - 6 lines] > astronomers know exactly what the practical limit of flare is for real > lenses; I don't.] Whether the lens is the limit or not depends on the scene. Take an extreme case where you have a very small bright spot that only fills say 8 pixels on a 8 MP sensor. Even if you spread all of this light across the rest of the pixels it would only be 1/1,000,000 of the level that is hitting the 8 pixels, which would be close to 20 stops. Scott > .com/imagedetail/raw.converter.shadow.detail > [quoted text clipped - 16 lines] > > Scott Scott, That reminds me of astrophotos. Here is an astrophoto that has 4th magnitude stars that are overexposed: http://www.clarkvision.com/astro/surface-brightness-profiles/m45 Very close to the brightest stars are measured nebulosity with brightnesses around 21 magnitudes/sq arc-second. Camera: canon 10D, 500 mm f/4 telephoto + 1.4x TC. That works out to 2.1 arc-seconds per pixel, so surface brightnesses are reduced by 1.6 magnitudes. Se we have a range of 21-1.6-4 = 15.4 stellar magnitudes, which is 2.512^15.4 = a dynamic range of 1,400,000 or 20.4 stops. The dark sty is even darker, so a stop or so more. So the lens has flair (the stars are overexposed), but a short distance away, the flair is down by over 21 stops! Roger >>Yes it is possible. Sunrises/sunsets are good examples. >>The sun gets completely blown and saturates the [quoted text clipped - 15 lines] > astronomers know exactly what the practical limit of flare is for real > lenses; I don't.] Scattered light in a lens system certainly limits dynamic range. But how much it limits depends on how close the bright subject is. Here is another test that shows just over 10 stops of dynamic range: http://www.clarkvision.com/imagedetail/dynamicrange See figure 6b. There are 6-stop changes in intensity that are close together. Of course, there might be greater ranges and lens flair reduced it to these levels. A more rigorous test here has 10 stops (verified) in one image: http://www.clarkvision.com/imagedetail/dynamicrange2 There are many very sharp boundaries, and little evidence for flair. When I get some time, I'll look into it more. Notice the slide film in figure 4 seems to show significant flair (scattering in the film, not the lens). Roger > If Canon split the sensor like Fuji did to > give a smaller "less sensitive" pixel beside each larger [quoted text clipped - 3 lines] > calibration of the different pixels is difficult, and > the benefits are small compared to the problems introduced. Hi Roger, Have you actually done any work with the Fuji sensor to get a look at its dynamics? I've seen no mention of problems being introduced by it. It certainly has limited the pixel growth potential of the camera for the reasons you mention above. Cheers, Alan Signature-- r.p.e.35mm user resource: http://www.aliasimages.com/rpe35mmur.htm -- r.p.d.slr-systems: http://www.aliasimages.com/rpdslrsysur.htm -- [SI] gallery & rulz: http://www.pbase.com/shootin -- e-meil: Remove FreeLunch. >> If Canon split the sensor like Fuji did to >> give a smaller "less sensitive" pixel beside each larger [quoted text clipped - 14 lines] > Cheers, > Alan Alan, No I have not worked with a Fuji. Some of the test data I've seen show a disconnect in intensity between the two on-chip pixel types, and there is a different slope to the characteristic curve: it has a kink in it. How much this translates into image detail and tonality I do not know, except when you put such effects into images with the Photoshop curves tool it doesn't look natural. Roger |
Reply to this Message |
 Sign In
 Join
 My Latest Posts My Monitored Threads My Blog My Photo Gallery My Profile My Homepage Start New Thread Enable EMail Alerts Rate this Thread
|
©2009 Advenet LLC Privacy Policy - Terms of Use
This website includes both content owned or controlled by Advenet as well as content owned or controlled by third parties.