 |
 | 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 / September 2005"Why Raw" Article
I have written Part I of an article titled "Why Raw". The article can be found on my web site at: http://ronbigelow.com/articles/raw/raw.htm Other articles can be found on my site at: http://ronbigelow.com/articles/articles.htm Comments are always welcome. Very informative. Thank you! >I have written Part I of an article titled "Why Raw". The article can > be found on my web site at: [quoted text clipped - 6 lines] > > Comments are always welcome. [] >> I have written Part I of an article titled "Why Raw". The article can >> be found on my web site at: >> >> http://ronbigelow.com/articles/raw/raw.htm [] >> Comments are always welcome. Ron, There is a major error in the article, section Raw Advantage #2: Bits, making the comparison between JPEG and RAW quite incorrect. The data in JPEG files has a gamma correction of (typically) 2.2 applied, which means that it can accommodate a far greater dynamic range than just 255 linear light levels. The difference between RAW and JPEG in is the accuracy of light level representation, not in the dynamic range. Cheers, David David, The 255 light levels have to do with tonal range not dynamic range. The article assumed a 5 stop dynamic range camera. That is fixed. The camera can not exceed that. The gamma curve can make the image more contrasty and make it appear to the eye that there is more dynamic range. However, that will simply readjust the spread of the 255 light levels (compressing some and spreading out others), but there will still only be no more than 255 light levels in a JPEG file. The article is correct. > David, > [quoted text clipped - 6 lines] > still only be no more than 255 light levels in a JPEG file. The > article is correct. The way I'm understanding it, "Figure 4: Shades vs. Stops of Light" for JPEG is wrong. When you half the exposure, while in the RAW file the digital value will be halved, this does not happen in JPEG. The values would be (approximately) 255, 186, 135, 99, 72, 52 etc. The JPEG has twenty different tonal values four stops down and five stops down from peak white, and 69 different tonal values between peak white and one stop down. You article gives 16 and 128 shades for these values. David >>David, >> [quoted text clipped - 16 lines] > > David Ron: Concerning your writeup at http://ronbigelow.com/articles/raw/raw.htm See Figure 7 at: Dynamic Range and Transfer Functions of Digital Images and Comparison to Film http://www.clarkvision.com/imagedetail/dynamicrange2 While in a jpeg file there can be no more than 255 levels, those levels do not have to be linearly distributed. In fact they are not, just as Figure 7 above shows. Note in that figure the dynamic range of the jpeg and the raw file cover the same range, except for a slight increase in noise at the low end. Of course, the 12-bit raw has more precision, and that is the main difference between the two. You might also check out: Digital Camera Raw versus Jpeg Conversion Losses http://www.clarkvision.com/imagedetail/raw.versus.jpeg1 The statements about digital camera dynamic range of the "better cameras" have about 5 stops is incorrect. The better cameras (e.g. DSLRs) are Poisson statistics limited (meaning photon statistics limited by the sensor's given quantum efficiency). The sensors have over 12-bits dynamic range and thus are limited by the 12-bit ADC. See Table 3 at: The Signal-to-Noise of Digital Camera images and Comparison to Film http://www.clarkvision.com/imagedetail/digital.signal.to.noise/ Roger >David, >The 255 light levels have to do with tonal range not dynamic range. The >article assumed a 5 stop dynamic range camera. That is fixed. The >camera can not exceed that. To say that a camera has 5 stops of dynamic range and no more assumes a certain quality standard. With a definition that gives 5 stops for DSLRs, you are talking about a standard of very high quality shadows. The fact is, there is no hard, concrete limit on the dynamic range of an image if your quality standards are flexible; the quality simply deteriorates as you go deeper and deeper into the shadows, until noise dominates and the signal is vague and lost in the chaos. >The gamma curve can make the image more >contrasty and make it appear to the eye that there is more dynamic >range. However, that will simply readjust the spread of the 255 light >levels (compressing some and spreading out others), but there will >still only be no more than 255 light levels in a JPEG file. The article >is correct. A JPEG can only express 256 light levels for each color channel. Most of the values less than about 50 only occur by demosaicing and sharpening when they come from a 12-bit RAW file. If you took each RAW color channel and made a bitmap from it in 8-bit gamma-adjusted space, with no sharpening or softening applied, most of the values below 50 would not be used at all. The histogram would be spaced like this: | | | | | | | |||||||||||| The section to the right, where the histogram is full (without gaps) is the range where the 12-bit RAW has more level-definition than the 8-bit gamma-corrected space. The rightmost section has several RAW values becoming a single 8-bit value.  Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< In message <1aiRe.99663$G8.11733@text.news.blueyonder.co.uk>, "David J Taylor" <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk.invalid> wrote: >There is a major error in the article, section Raw Advantage #2: Bits, >making the comparison between JPEG and RAW quite incorrect. The data in >JPEG files has a gamma correction of (typically) 2.2 applied, which means >that it can accommodate a far greater dynamic range than just 255 linear >light levels. The difference between RAW and JPEG in is the accuracy of >light level representation, not in the dynamic range. An 8-bit TIFF has more potential shadow detail than a 12-bit RAW file. It is not usually realized, however, as 8-bit TIFFs generally come from 12-bit RAW files. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< [] > An 8-bit TIFF has more potential shadow detail than a 12-bit RAW file. > It is not usually realized, however, as 8-bit TIFFs generally come > from 12-bit RAW files. You mean that the 8-bit TIFF is gamma-corrected as well as the JPEG, and that at the range in the TIFF from lowest brightness levels to the highest exceeds that of a 12-bit RAW file? 12-bit - 4096 linear levels from 0 to 4095 - range 4095:1 8-bit gamma corrected, 255 levels from 5E-6 to 1 (or 0.001 to 255) - range 200,000:1 David > [] >> An 8-bit TIFF has more potential shadow detail than a 12-bit RAW [quoted text clipped - 9 lines] > 8-bit gamma corrected, 255 levels from 5E-6 to 1 (or 0.001 to 255) - > range 200,000:1 The problem is that the TIFF comes from the same 12 bit source. -Mike >> [] >>> An 8-bit TIFF has more potential shadow detail than a 12-bit RAW [quoted text clipped - 13 lines] > > -Mike Yes, I appreciate that. I just wanted to confirm precisely what John was saying, as it's important that the information on Ron's page is correct. David In message <MFxRe.100197$G8.94336@text.news.blueyonder.co.uk>, "David J Taylor" <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk.invalid> wrote: >You mean that the 8-bit TIFF is gamma-corrected as well as the JPEG, and >that at the range in the TIFF from lowest brightness levels to the highest >exceeds that of a 12-bit RAW file? From what I've seen and heard, a TIFF is more likely to be literal in it's 2.2 gamma in the shadows, whereas a JPEG is more likely to render the deeper shadows even deeper, to hide noise. >12-bit - 4096 linear levels from 0 to 4095 - range 4095:1 Yes, literally; level 4095 is 4095 times as bright as level 1. >8-bit gamma corrected, 255 levels from 5E-6 to 1 (or 0.001 to 255) - range >200,000:1 In therms of luminance, it is actually (255^2.2):1 or 196,965:1. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > In message <MFxRe.100197$G8.94336@text.news.blueyonder.co.uk>, > "David J Taylor" [quoted text clipped - 17 lines] > > In therms of luminance, it is actually (255^2.2):1 or 196,965:1. Thanks, John. Our understanding is the same, then. I only quoted 200,000:1 simply because if I quoted the exact number I expected someone to think I was being pedantic! I can appreciate that JPEG might be optimised in a slightly different way to TIFF so that errors are masked. By the way (for anyone else who's still reading), this 2.2 gamma figure is what some cameras call the "contrast" setting - use a smaller number when converting from linear to JPEG and you'll get more contrast, a larger number gives less contrast. David In message <0KSRe.100820$G8.39510@text.news.blueyonder.co.uk>, "David J Taylor" <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk.invalid> wrote: >> In therms of luminance, it is actually (255^2.2):1 or 196,965:1. >Thanks, John. Our understanding is the same, then. I only quoted >200,000:1 Sorry; I didn't even notice that you wrote that. I read it fast, and somehow thought you were writing about 255:1. >simply because if I quoted the exact number I expected someone >to think I was being pedantic! I can appreciate that JPEG might be >optimised in a slightly different way to TIFF so that errors are masked. >By the way (for anyone else who's still reading), this 2.2 gamma figure is >what some cameras call the "contrast" setting - use a smaller number when >converting from linear to JPEG and you'll get more contrast, a larger >number gives less contrast. Most JPEG conversions, I think, vary the gamma across the histogram. This is important if you want to keep middle grey standard, but expand or compress contrast relative to it. I really wish I had a greyscale wedge the quality of the Gretag-MacBeth color checker, but with about 10 stops of grey rectangles with steps of about 1/6 stop, or even a pie wedge to discount light roll-off. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< [] > Most JPEG conversions, I think, vary the gamma across the histogram. > This is important if you want to keep middle grey standard, but expand > or compress contrast relative to it. Yes, there may also be an attempt to keep a little of the specular highlights, for example. > I really wish I had a greyscale wedge the quality of the > Gretag-MacBeth color checker, but with about 10 stops of grey > rectangles with steps of about 1/6 stop, or even a pie wedge to > discount light roll-off. Could you make something out of a light box and some grey film in multiple layers? I guess that wouldn't be accurate enough, though. David > In message <0KSRe.100820$G8.39510@text.news.blueyonder.co.uk>, > "David J Taylor" [quoted text clipped - 30 lines] > John P Sheehy <JPS@no.komm> > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< Re the grayscale you desire ... get a Kodak Q-13 "Kodak Color Separation Guide and Gray Scale-Small," costs something like US$20 at places such as B&H. http://www.bhphotovideo.com/bnh/controller/home?ci=1&sb=ps&pn=1&sq=desc&Ini tialSearch=yes&O=SearchBar&A=search&Q=*&bhs=t&shs=Kodak+Q-13&image.x=10&ima ge.y=5 They are literally an "industry standard," and have been used as such for over fifty years. SNIP > Re the grayscale you desire ... get a Kodak Q-13 "Kodak Color > Separation Guide and Gray Scale-Small," costs something like > US$20 at places such as B&H. Unfortunately the range they cover is too small for a decent DSLR. The one you'll need is e.g. the Stouffer T4110: http://www.stouffer.net/TransPage.htm#transmission%20step It covers a range of approx. 10000:1 or 13.5 stops. That should be enough to be able to determine the dynamic range with a single exposure. Bart >SNIP >> Re the grayscale you desire ... get a Kodak Q-13 "Kodak Color [quoted text clipped - 9 lines] >enough to be able to determine the dynamic range with a single >exposure. That looks interesting. Do you have one? Is glare an issue? Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >Is glare an issue? I guess not, as long as they are illuminated from behind; they are transmission wedges. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< >> Is glare an issue? > > I guess not, as long as they are illuminated from behind; they are > transmission wedges. The company seems to make both transmission and reflection wedges, quite reasonably priced (I would have said) if you don't want hand calibration. http://www.stouffer.net/ David >>Is glare an issue? > > I guess not, as long as they are illuminated from behind; they are > transmission wedges. Actually to get the best results, one should *only* have back-lighting for such dense areas as the target offers. Any light striking it from the front will reduce contrast a little. Of course, if the amount of backlighting is much larger than any ambient light to the front, there is no issue. That means that it's best to shoot the target with the front shielded from ambient light, and the back only lighted by a lightsource that's shielded from lighting anything else than the target. Bart > In message <431a2d19$0$11061$e4fe514c@news.xs4all.nl>, SNIP > That looks interesting. Do you have one? Is glare an issue? I have one from Kodak that covers 10 stops, but that is too limited because the channels are not equally sensitive, which requires combining different exposures to clip on the bright side, and be noise limited by dynamic range on the dark side. I'm in the process of ordering the T4110, and I'm building a dedicated lightbox/contraption with masked opaline glass as a diffusor, and a kind of ambient-light shade on the front. The Imatest software I use will also allow to make sure that the illumination of the target is uniform, and I can measure/calibrate the target (in place) transmission and spectral characteristics with the Eyeone Photo's Spectrophotometer. I'm experimenting with the lightsource to use, but I'll probably either end up with 5300K or 6500K fluorescent tube lighting (which makes it hard to restrict lighting to only the target holder), or a shielded electronic flash rear window. It'll probably be the flash exposure (more consistent/repeatable and shorter exposure times). Bart >Re the grayscale you desire ... get a Kodak Q-13 "Kodak Color Separation >Guide and Gray Scale-Small," costs something like US$20 at places such as [quoted text clipped - 6 lines] >They are literally an "industry standard," and have been used as such for >over fifty years. I already have this, and the quality is poor. It is glossy, and they grey rectangles are reddish. You have to use special lighting for the glare not to be an issue. Signature<>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS@no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< Tnx for article, it's great. I'm reading the rest of your work... >I have written Part I of an article titled "Why Raw". The article can > be found on my web site at: [quoted text clipped - 6 lines] > > Comments are always welcome. > I have written Part I of an article titled "Why Raw". The article can > be found on my web site at: [quoted text clipped - 6 lines] > > Comments are always welcome. Thanks again. I only have one comment: It would be cool if you made your articles in an easy to print format, like pdf. I wanted to read it this morning, but was running a little late.... I could've printed it and read it during my lunch break. -- Slack >I only have one comment: It would be cool if you made your articles in >an easy to print format, like pdf. I wanted to read it this morning, >but was running a little late.... I could've printed it and read it >during my lunch break. If you are so unhappy, why don't you ask Ron for a refund? ;-) Tony Polson wrote: >> I only have one comment: It would be cool if you made your articles >> in an easy to print format, like pdf. I wanted to read it this [quoted text clipped - 4 lines] > > ;-) He's such a nice bloke, I think he would give double your money back. :-) -Mike > Tony Polson wrote: > [quoted text clipped - 11 lines] > > -Mike I'm such a valued customer, he gave triple my $$ back. So there :-D -- Slack >I only have one comment: It would be cool if you made your articles in >an easy to print format, like pdf. I wanted to read it this morning, >but was running a little late.... I could've printed it and read it >during my lunch break. FWIW, some newer printers have a util to help print web pages. What's the problems with the pages that you couldn't print them? Too wide, choose landscape. Otherwise you can use Adobe Acrobat to print through Distiller to PDF. if you want something like this. ---------- Ed Ruf Lifetime AMA# 344007 (Usenet@EdwardG.Ruf.com) See images taken with my CP-990/5700 & D70 at http://edwardgruf.com/Digital_Photography/General/index.html >>I only have one comment: It would be cool if you made your articles in >>an easy to print format, like pdf. I wanted to read it this morning, [quoted text clipped - 7 lines] > ---------- > Ed Ruf Lifetime AMA# 344007 (Usenet@EdwardG.Ruf.com) Actually, that is what I tried to do and my printer is brand spanking new (Dell 3100cn). Some of the graphics were getting split between two different pages. Some of the gray side bars in the 2nd half of the article got all jacked up: overlapping with the main text. I tired playing around resizing the print, but no joy. I tried to print via two web browsers, FF & Mozilla.... maybe that was the problem?? What is Distiller? I'm always willing to learn new stuff. -- Slack > Actually, that is what I tried to do and my printer is brand spanking > new (Dell 3100cn). you're printer driver should have a print-to-fit option. it's fairly common. else, some of the alternative browsers will actually resize the page graphics and all (not just the font like IE). SignatureMark Photos, Ideas & Opinions http://www.marklauter.com [] > What is Distiller? I'm always willing to learn new stuff. http://www.adobe.com/products/acrdis/main.html >[] > [quoted text clipped - 6 lines] > > Open Office can export to PDF and its free! A |
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.