Luminous Landscape Forum
Raw & Post Processing, Printing => Adobe Lightroom Q&A => Topic started by: Marshallarts on March 02, 2010, 11:47:55 am
-
From what I understood, bit depth also equals the amount of stops of dynamic range you are able to effectively have. An 8bit jpg only has 8 stops, a 16bit Tiff will have 16 stops. The bottle neck of course being the camera's limitations to how much DR is can capture. But I just did a test and am a little confused,
I took a raw cr2 file and changed exposure to -4ev and exported as a 16-bit Tiff. I did the same changing exposure to +4. When I opened up the files in Photoshop where I am able to adjust exposure under the adjustments tab. Brining back the -4 and +4 exports to 0ev and comparing it to the original looks completely different. The -4ev one performed nicely but has a grayish feel to it like the contrast is down. The +4ev on, however, was completely blown out and unable to be recovered. I thought saving as 16bit Tiff would allow for the dynamic range to be preserved so I could recover it?
I did the same only saved -4 and +4 as DNG and this time importing them in Photoshop and adjusting in ACR back to 0ev they were identical to the original 0ev image.
If Tiff is 16-bits why could it not preserve the dynamic range like the DNG could? When I exported the Tiffs did it throw away the information that I was clipping on my histogram? I thought part of the reason we do 16bits is to preserve that info.
Look forward to hearing your feedback.
-
I don't think there is a direct 1:1 correlation between bit depth and dynamic range, although a 16-bit file (really, it's only 12 or 14 bits) does have more dynamic range than an 8-bit file.
As for clipping, it depends on which histogram you're talking about. The histogram on the camera is based on the JPEG preview (8-bit), so you will be able to recover some of the clipped areas during conversion as a RAW has more dynamic range, but it really depends on how much was actually clipped. The histogram in LR or ACR will be based on the actual RAW data, so it's more accurate. If you cannot recover clipped data with the highlight or shadow sliders, you're essentially out of luck. RAW isn't a magic solution and does need some attention to detail during exposure.
A +4 EV exposure will more than likely have some areas of unrecoverable clipping
-
From what I understood, bit depth also equals the amount of stops of dynamic range you are able to effectively have. An 8bit jpg only has 8 stops, a 16bit Tiff will have 16 stops. The bottle neck of course being the camera's limitations to how much DR is can captur
Wrong, wrong wrong...
Bit depth and dynamic range have NOTHING to do with one another...seriously, whoever told you that is full of doodoo...
Bit depth is related to color space in terms of the levels distribution of the color and tone through out the color space...
And TIFF is normally locked into a gamma encoded color space while original raw captures are in linear gamma with the cameras' spectral response as color data.
Once you take that raw file and process it into a gamma encoded color space the distribution of tone is no longer linear so all the data that is clumped near the highlights has been redistributed. The will severely limit the amount of highlight recovery you can to to a processed TIFF file...
The dynamic range of your camera sensor is another thing entirely...you determine that based on the point where the sensor floods (clips) compared to the point at which noise becomes objectionable. Most sensors are their "native" ISO these days will be between 10-12 stops...some medium format backs will be higher and some point and shoot cameras will be lower. But the dynamic range of your digital capture has NOTHING to do with the file format bit depth you are using...
-
thanks for clearing up the confusing. My basis for saying bit depth relates to dynamic range comes from the following paragraph pulled from "HDR Photography Photo Workshop 2009" by Pete Carr and Robert Correll:
"The bit depth of a camera is analogous to the stop scale. Each stop doubles or halves light, which is exactly how bits work. Adding a bit to a binary number doubles it. Conversely, removing the leading bit of a binary number halves it. Each bit, therefore, provides a stop’s worth of dynamic range. Thus, a 12-bit camera has 12 stops of potential dynamic range, which results in 4096 levels of light-to-dark discrimination in each color channel."
He then goes on to address what your brought up about color, "Finally, bit depth can also be used to describe the number of colors a camera can capture and store in a file. In this sense, a camera with a greater bit depth has more color-sensing ability. Each color channel of a 12-bit camera can contain up to 4096 different shades of that color, resulting in a tremendous number of total colors that the three channels can combine to reproduce. However, dynamic range is the story of brightness, not color. As mentioned previously (interestingly he never mentioned anything previously), the shades of color are not evenly distributed across the bit- depth system, which results in a skewed intensity range for each color. Therefore, having 4096 pos- sible shades of each color is not enough to cap- ture the true dynamic range of a scene without compromise. If it were, you would never need worry about a blue sky turning white."
ckimmerle -- I was referring to looking at the histogram on Lightroom. The Photo was exposed normally and fit properly within the histogram and minimal clipping. I intentionally raised EV +4 which made everything very blown out and on the right side of the histogram in Lightroom with much clipping. I saved it as a 16-bit Tiff and attempted to recover in Photoshop. This entire test was to see if in fact the 16bit Tiff preserved that information since what I read claims a 16bit Tiff has 16 stops of range. If my monitor is showing me only 8 stops (8bits) then pushing the photo +4 and saving as 16bit Tiff I theoretically should be able to recover it back to normal. This depends on where the histogram lays within the 16 stops of DR. If in fact what we see on the histogram is identical to the 8-bits we see in a jpeg if the histogram was located in the center of the 16 stops I should have 4 stops on either side. But because my test failed and I was unable to recover the photo it makes me wonder if the histogram lays on the right side of that DR. Meaning if what I read is true about 16bit Tiff having 16 stops of DR then those would exist on the shadow end. My -4EV experiment faired much better but lost detail and contrast. The DNGs I tried worked flawlessly.
I believe my results have to do with what Schewe suggested about the gamma being applied to the Tiff files. Pushing +4EV moved my detail off the right and effectively lost the information in clipping.
Am I correct?
-
bit depth relates to dynamic range
Your camera likely produces 12-bit RAW files (some do 14-bit like my Nikon D700) when the ADC converts the analog signal data into digital information.
The dynamic range of the camera must fit within this limitation, but does not necessarily "fill" it completely.
I intentionally raised EV +4 which made everything very blown out and on the right side of the histogram in Lightroom with much clipping. I saved it as a 16-bit Tiff and attempted to recover in Photoshop.
By forcing the image to clip when you rendered it, you threw away TONS of data that is unrecoverable.
Am I correct?
Not really ... but you'll figure it out ...
Read this website beginning to end ... http://www.cambridgeincolour.com/tutorials.htm (http://www.cambridgeincolour.com/tutorials.htm)
-
thanks for clearing up the confusing. My basis for saying bit depth relates to dynamic range comes from the following paragraph pulled from "HDR Photography Photo Workshop 2009" by Pete Carr and Robert Correll:
"The bit depth of a camera is analogous to the stop scale. Each stop doubles or halves light, which is exactly how bits work. Adding a bit to a binary number doubles it. Conversely, removing the leading bit of a binary number halves it. Each bit, therefore, provides a stop’s worth of dynamic range. Thus, a 12-bit camera has 12 stops of potential dynamic range, which results in 4096 levels of light-to-dark discrimination in each color channel."
The way I look at is is this (I hope someone will correct me if I am wrong). In an 8 bit file, each of the 3 colors is represented by 8 bits of data representing 256 different values of 0 through 255. In a 16 bit file, each color is represented by 16 bits of data representing 64,000+ levels - but still 0 through 255. Thus, 16 bit lets you represents finer gradations of color within the same overall range but it does not let you represent a greater range - black is not any blacker and white is not any whiter.
-
Wrong, wrong wrong...
Bit depth and dynamic range have NOTHING to do with one another...seriously, whoever told you that is full of doodoo...
Bit depth is related to color space in terms of the levels distribution of the color and tone through out the color space...
And TIFF is normally locked into a gamma encoded color space while original raw captures are in linear gamma with the cameras' spectral response as color data.
Not entirely wrong! In a linear raw file, the dynamic range is f/stops can be no greater than the bit depth as Sean McHugh explains in his excellent tutorial on dynamic range Cambridge in Color (http://www.cambridgeincolour.com/tutorials/dynamic-range.htm). Norman Koren (http://www.normankoren.com/digital_tonality.html) elaborates further (see his exposure zone chart). In an 8 bit raw file, the maximum value would be 1024 and the minimum value would be 1; the DR would be 1024/1 or 8 stops. That would allow only one level in the darkest f/stop, but to be useful that last f/stop would require several levels so the DR would be even less.
Gamma encoding compresses redundant highlight tones and allows more room for shadow values, increasing the number of f/stops than be encoded with a given bit level as Norman's chart shows.
This discussion assumes that there is no tone mapping of the DR. In practice, an S-curve or other adjustments would be applied to the data and an 8 bit JPEG can represent more than 8 stops of dynamic range in the scene. See Karl Lang's (http://www.adobe.com/digitalimag/ps_pro_primers.html) excellent white paper on rendering of the print and also Norman's discussion.
Once you take that raw file and process it into a gamma encoded color space the distribution of tone is no longer linear so all the data that is clumped near the highlights has been redistributed. The will severely limit the amount of highlight recovery you can to to a processed TIFF file...
Highlight recovery in raw is possible because the color channels are not equally sensitive in digital cameras and the green channel can be blown while the red and blue channels contain data. Once the file is white balanced for output as a TIFF this recovery is no longer possible. Still in a 16 bit TIFF there is plenty of highlight detail as Norman explains in his tutorial. After all, the human visual system can detect only about 100 levels at most in the brightest f/stop of a digital file. An 16 bit gamma 2.2 encoded file contains 17,712 levels in the brightest f/stop.
The dynamic range of your camera sensor is another thing entirely...you determine that based on the point where the sensor floods (clips) compared to the point at which noise becomes objectionable. Most sensors are their "native" ISO these days will be between 10-12 stops...some medium format backs will be higher and some point and shoot cameras will be lower. But the dynamic range of your digital capture has NOTHING to do with the file format bit depth you are using...
This is largely true since most digital sensors are limited more by shadow noise than the possible number of encoded levels. Emil Martinec (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p2.html#SNR-DR) gives a good analysis and points out that the high end Nikons and Canons offering 14 bit raw files don't have enough true DR to justify 14 bits and the extra bit depth is wasted on encoding noise.
-
The way I look at is is this (I hope someone will correct me if I am wrong). In an 8 bit file, each of the 3 colors is represented by 8 bits of data representing 256 different values of 0 through 255. In a 16 bit file, each color is represented by 16 bits of data representing 64,000+ levels - but still 0 through 255. Thus, 16 bit lets you represents finer gradations of color within the same overall range but it does not let you represent a greater range - black is not any blacker and white is not any whiter.
Close ...
http://www.normankoren.com/digital_tonality.html (http://www.normankoren.com/digital_tonality.html)
EDIT ... sorry, didn't realize this link was in the preceding post ...
-
Jeff's comment above is dead on. I would add that a RAW file isn't really a picture yet - it is a collection of ones and zeroes. It has no color depth, color space, curves, or anything else "applied" or "assigned" to it. It becomes a picture when you process it.
As soon as you process a RAW file (for example, open in Photoshop) all kinds of interesting things happen to it. Remember that the original RAW file is preserved, unless you deliberately delete it.
One tip: remember to try and keep your important image files in 16-bit as long as you can. Images tolerate editing better in 16-bit and a larger color space, such as Adobe 98 or Pro Photo RGB.
David
-
In a 16 bit file, each color is represented by 16 bits of data representing 64,000+ levels - but still 0 through 255. Thus, 16 bit lets you represents finer gradations of color within the same overall range but it does not let you represent a greater range - black is not any blacker and white is not any whiter.
Uh no..actually Photoshop doesn't actually work in 16 bits per channel...it works in 15 bits plus one level so the total max in Photoshop is 0 (zero) through 32768 for a total of 32769 levels in a Photoshop "16 bit" image (I call them high bit-depth). There are image processing reasons why the engineers decided to do this and the functional difference between a 15+1 bit file and a full 16 bit file is minimal (unless you are working with CGI images and masks)...
But then your statement that the levels are "still 0 through 255" is the wrong way to think about it. You are right that 0 (zero) is black and white is the lightest you can go, which in an 8 bit/channel file would be 255 but in the high bit-depth file the white is literally level 32768.
Then of course you can 32 bit floating point image data where black can be darker than black and white, whiter than white...(well, more or less)...
-
Jeff's comment above is dead on. I would add that a RAW file isn't really a picture yet - it is a collection of ones and zeroes. It has no color depth, color space, curves, or anything else "applied" or "assigned" to it. It becomes a picture when you process it.
Agreed!
-
Jeff's comment above is dead on. I would add that a RAW file isn't really a picture yet - it is a collection of ones and zeroes. It has no color depth, color space, curves, or anything else "applied" or "assigned" to it. It becomes a picture when you process it.
As soon as you process a RAW file (for example, open in Photoshop) all kinds of interesting things happen to it. Remember that the original RAW file is preserved, unless you deliberately delete it.
David
To extend this analysis even further, I would say that a 16 bit ProPhotoRGB tiff isn't a picture until you print it or view it on the screen. Prior to that, it is also a collections of ones and zeros just like a raw file. A raw file can be viewed without demosaicing. It contains color information just the same as the TIFF, but in a different format. Rather than in separate RGB layers which are themselves monochrome, the color information is in mosaic form. The color of the pixels in the raw file is defined by the response of the individual Bayer array pixels as seen through a color filter. The attachment illustrates this. The demosaiced file is shown on top and the raw file is on the bottom as viewed with Rawnalize.
[attachment=20611:raw_tiff.png]
-
I took a raw cr2 file and changed exposure to -4ev and exported as a 16-bit Tiff. I did the same changing exposure to +4. When I opened up the files in Photoshop where I am able to adjust exposure under the adjustments tab. Brining back the -4 and +4 exports to 0ev and comparing it to the original looks completely different. The -4ev one performed nicely but has a grayish feel to it like the contrast is down. The +4ev on, however, was completely blown out and unable to be recovered. I thought saving as 16bit Tiff would allow for the dynamic range to be preserved so I could recover it?
I did the same only saved -4 and +4 as DNG and this time importing them in Photoshop and adjusting in ACR back to 0ev they were identical to the original 0ev image.
If Tiff is 16-bits why could it not preserve the dynamic range like the DNG could? When I exported the Tiffs did it throw away the information that I was clipping on my histogram? I thought part of the reason we do 16bits is to preserve that info.
I think some of you guys in your replies might have misunderstood the question.
My experience is, if the image is adjusted in ACR so that no clipping of either shadows or highlights is apparent in the histogram, then a 16 bit TIFF conversion of the RAW file will essentially preserve the full DR and tonality.
This is essentially like asking, is there any advantage in making as many adjustments as possible in ACR before converting to TIFF, or could one do an equally good job making a straight conversion to TIFF with no adjustments other than to ensure there's no clipping of shadows and highlights, then adjust the image in Photoshop to taste?
There may be some advantage in doing as many adjustments as possible in ACR before conversion, but my impressions is they are rather subtle.
One technique that used to be promoted by Adobe to extract the maximum DR from an image, was to create two artificial exposures in ACR of the same image, using say +2 EV for one conversion (to maximise the shadow detail), and -2 EV for the second conversion (to maximise the highlight detail), then combine the two images in Photoshop. I used to employ this technique myself and thought I was getting a substantial improvement.
One day I took the trouble to do a comparison between the 'merged to HDR technique' and a single 16 bit TIFF of the -2EV conversion adjusted in PS using whatever skills I possess.
My impression was, that using the right technique, the 16 bit Tiff conversion at -2EV contained all the image information I needed to get essentially the same degree of DR and tonality that I could get from the other technique of merging the -2EV with the +2EV conversions.
I got the impression that the merging of the two artificial exposures was merely another way of skinning the cat.
However, I don't want to imply that having converted one's RAW images to Tiff, one can then dump the RAW images. If one did that one would then be depriving oneself of the benefits of any future improvements in RAW converters as well as the benefits of existing alternative RAW converters.
-
My experience is, if the image is adjusted in ACR so that no clipping of either shadows or highlights is apparent in the histogram, then a 16 bit TIFF conversion of the RAW file will essentially preserve the full DR and tonality.
No, actually I don't think it will...
If you look at the extreme amount of data in the top 1 stop of exposures (in either a 12 or 14 bit capture) the density of that data while in linear gamma is NOTHING like the amount of data once you clamp the data to a gamma encoded color space. I don't care how good you are at Photoshop curves, once that linear data is encoded into a non-linear gamma you'll NEVER be able to tease that data apart...
Will it matter to the vast majority of images" prolly not...but I have images whose extreme highlight data was teased way down to the deep shadows (an extreme case of ETTR) where if that image had been gamma encoded, no amount of teasing would have recovered the highlight detail.
I would also point out I used to do a lot of dual (or multiple of 3 or more) process blends in Camera Raw prior to Camera Raw 5.x primarily because we had no capability to tame a tone curve locally...now we do and the number of times I've had to resort to an blend of ACR processes has dropped considerably...
Which really is an even GREATER reason not to process raw images to TIFF with the expectation that the TIFF will stand up to later technical scrutiny...now more than ever in photography we are faced with the odd situation where in the future (near future with some products) the newly processed raw files will be substantially better than in the past...I seriously wouldn't want to be stuck dealing with pre-processed TIFF files instead of raw files.
-
Hi,
Add to that file size is much larger than with raw or DNG. Two reasons, demosaic process triples the amount of data. Tiff is 16 bit while DNG data is probably 12 bit or 14 bit at most. 16 bit TIFF files are big, like 128 MByte although the opriginal data is only perhaps 35 MByte.
Best regards
Erik
No, actually I don't think it will...
If you look at the extreme amount of data in the top 1 stop of exposures (in either a 12 or 14 bit capture) the density of that data while in linear gamma is NOTHING like the amount of data once you clamp the data to a gamma encoded color space. I don't care how good you are at Photoshop curves, once that linear data is encoded into a non-linear gamma you'll NEVER be able to tease that data apart...
Will it matter to the vast majority of images" prolly not...but I have images whose extreme highlight data was teased way down to the deep shadows (an extreme case of ETTR) where if that image had been gamma encoded, no amount of teasing would have recovered the highlight detail.
I would also point out I used to do a lot of dual (or multiple of 3 or more) process blends in Camera Raw prior to Camera Raw 5.x primarily because we had no capability to tame a tone curve locally...now we do and the number of times I've had to resort to an blend of ACR processes has dropped considerably...
Which really is an even GREATER reason not to process raw images to TIFF with the expectation that the TIFF will stand up to later technical scrutiny...now more than ever in photography we are faced with the odd situation where in the future (near future with some products) the newly processed raw files will be substantially better than in the past...I seriously wouldn't want to be stuck dealing with pre-processed TIFF files instead of raw files.
-
No, actually I don't think it will...
If you look at the extreme amount of data in the top 1 stop of exposures (in either a 12 or 14 bit capture) the density of that data while in linear gamma is NOTHING like the amount of data once you clamp the data to a gamma encoded color space. I don't care how good you are at Photoshop curves, once that linear data is encoded into a non-linear gamma you'll NEVER be able to tease that data apart...
That reasoning is specious. Of the 8192 possible levels in the brightest f/stop of a 14 bit raw file, most are noise. Remember, shot noise is greatest in the highlights since it varies as the square root of the number of photons captured. Here some histograms of Nikon D3 flatfields near saturation. The upper field is from only 1 exposure and the noise includes fixed pattern noise as well as shot noise. The bottom is a subtraction of two identical exposures to get the random noise, which is almost entirely shot noise. Since the standard deviation of the latter is 56, the granularity of the measurement is much greater than one level.
[attachment=20616:Sat_200_..._plus300.png]
Since the brighter f/stops of raw files contain redundant levels, Nikon lossy NEF compression discards most of them with virtually no loss of image quality (see Emil Martinec (http://theory.uchicago.edu/~ejm/pix/20d/tests/noise/noise-p3.html#bitdepth)). Those brightest f/stops are not so information rich after all.
Furthermore, the number of levels in the brightest f/stop of a 16 bit gamma 2.2 encoded file is 17,712 (see Norman Koren reference in my previous post). Another way to look at the number of levels in a 12 bit gamma 1.0 file when converted to a 15 bit Photoshop file is to use Bruce Lindbloom's levels calculator. The number of levels are the same! The 15 bit file contains all the levels!
[attachment=20617:Levels.gif]
I rest my case. Comments are welcome, but I expect none from the forum Rottweiler.
-
I rest my case. Comments are welcome, but I expect none from the forum Rottweiler.
It would be fun to see it go after the forum Chihuahua <g>
-
It would be fun to see it go after the forum Chihuahua <g>
+1
-
Will it matter to the vast majority of images" prolly not...but I have images whose extreme highlight data was teased way down to the deep shadows (an extreme case of ETTR) where if that image had been gamma encoded, no amount of teasing would have recovered the highlight detail.
Here is the image I was referring to that proves that the brightest stop of a linear capture contains a lot of useful (and usable) data...perhaps a real image is more useful than theory?
(http://schewephoto.com/test/default.jpg)
This is the default image inside Camera Raw showing the histogram far, far to the right.
(http://schewephoto.com/test/adjusted.jpg)
Here is the result of Camera Raw adjustments which ironically only used a -0.08 Exposure and Recovery of 45. The Brightness was at -113 and the Blacks pinned at 100 to draw the upper data down. In the Parametric Curves Shadows were also set to -100 while in the Point Curve editor the upper tonality was teased apart...
So, yes, I think it can be said the brightest stop of data in a raw capture has a ton of useful image data...no?
:~)
P.S. and before you ask, the shot was done at F22 the max F stop of my 24-105 lens at 1/2 second...I did a bracket trying to capture the water in motion...when I looked at the LCD of course the preview looked like the unadjusted ACR preview. I thought the shot was a lost cause but that brightest stop of data still had a lot of useful image data in it...the image looks pretty nice as a 24" x 30" print...
-
(http://schewephoto.com/test/adjusted.jpg)
the shot was done at F22 the max F stop of my 24-105 lens at 1/2 second...I did a bracket trying to capture the water in motion...when I looked at the LCD of course the preview looked like the unadjusted ACR preview. I thought the shot was a lost cause but that brightest stop of data still had a lot of useful image data in it...the image looks pretty nice as a 24" x 30" print...
Fascinating!
while in the Point Curve editor the upper tonality was teased apart...
What does that mean and how did you do that?
-
Thanks for all the incredibly detailed responses! It's going to take me awhile to digest all of this. And thank you for kindly providing links, I look forward to referencing them time and time again in the future.
I found another quote that helped me also. It's from Mastering HDR Photography by Michael Freeman.
"To clear up any confusion about dynamic range and bit depth, the number of bits determines the precision of the capture--how many steps the range of brightness can be recorded as. In other words, a 14-bit sensor is more accurate than a 12-bit sensor. Though it might, it does not necessarily capture a wider range, but more detail within the range. The real limit to the range of a sensor is the amount of noise in the shadows."
I guess my earlier quote from the other book says "potential dynamic range" not that bit depth = dynamic range, an oversight of mine to assume it did.
Although my mind isn't completely wrapped around the info you all possess, and are taking sides on, I feel there is much to learn from both sides. Keeping the original raw files certainly makes the most sense as they are smaller and preserve the original data. My question was provoked not as a study of dynamic range in tiff vs raw, but in recovery. By deliberately pushing EV either + or - so that your image looks bad (the furthest PS or LR allows is -4 or +4), after saving as a tiff or raw file could you recover it to the original look. My thought at first was maybe you could. This all depends on how much DR can be preserved in these files and if the range is preserved equally on both sides to determine if this will work with both +4 and -4 effected images.
After looking at my test I realize the Exposure tool within PS is different than Exposure in Lightroom. Though opening the +4EV DNG in ACR I can easily move the exposure slider back down and get my original image, if I were to open it in Photoshop without touching it in ACR and use the Exposure tool in Photoshop it appears the same way the Tiff looks when I try to recover using the Exposure tool in Photoshop. So I guess my test is flawed.
There is a reason why I'm experimenting with this. I create time lapses. In situation such as sunsets it's very hard to photograph because you have to adjust your exposure every few minutes for the changing light conditions. This makes terrible exposure jumps when you create a image sequence, unusable. I've found using Lightroom you can "match total exposures" to smooth out the jumps. But in the case of a sunset as it gets darker and you've compensated by increasing shutter speed, after you "match total" exposures LR will apply drastic -EV to compensate for the increased shutter speed. I was hoping by saving as a 16bit Tiff or DNG that info would be preserved so that I could use After Effects to create a sequence and apply a gradual exposure increase throughout.
What do you guys think?
-
"To clear up any confusion about dynamic range and bit depth, the number of bits determines the precision of the capture--how many steps the range of brightness can be recorded as. In other words, a 14-bit sensor is more accurate than a 12-bit sensor. Though it might, it does not necessarily capture a wider range, but more detail within the range. The real limit to the range of a sensor is the amount of noise in the shadows."
The old analogy that still holds up (well for those of us that agree with the above) is this: Dynamic range is akin to the distance in a staircase. Bit depth is the number of steps in the staircase. You can have a staircase that is 20 feet from bottom to top. One could have 20 steps, one could have 40 steps. The distance (dynamic range) is the same. More steps doesn’t alter the height of the staircase.
-
And TIFF is normally locked into a gamma encoded color space while original raw captures are in linear gamma with the cameras' spectral response as color data.
Once you take that raw file and process it into a gamma encoded color space the distribution of tone is no longer linear so all the data that is clumped near the highlights has been redistributed. The will severely limit the amount of highlight recovery you can to to a processed TIFF file...
***********
Over my head but from a "users" perspective, am I better off exporting 16 bit TIFF or DNG from Phocus and then to import into LR or PS?
Steve
-
Here is the image I was referring to that proves that the brightest stop of a linear capture contains a lot of useful (and usable) data...perhaps a real image is more useful than theory?
This is the default image inside Camera Raw showing the histogram far, far to the right.
Here is the result of Camera Raw adjustments which ironically only used a -0.08 Exposure and Recovery of 45. The Brightness was at -113 and the Blacks pinned at 100 to draw the upper data down. In the Parametric Curves Shadows were also set to -100 while in the Point Curve editor the upper tonality was teased apart...
So, yes, I think it can be said the brightest stop of data in a raw capture has a ton of useful image data...no?
Yes, there are useful data in that brightest f/stop, but Ray stipulated that there should be no clipping in the image prior to its rendering into a 16 bit TIFF. Your image appears strongly clipped and violates Ray's stipulation and also my prior stipulation. It remains to be shown that with non clipped files RAW is superior to a 16 bit TIFF (or PS 15+1). There are editors other than Photoshop.
Highlight recovery could not have been performed in a white balanced TIFF. Also, have you looked at the actual raw file to see how much clipping is really present?. You can't necessarily rely on the camera or ACR histogram to reflect the raw values.
-
The old analogy that still holds up (well for those of us that agree with the above) is this: Dynamic range is akin to the distance in a staircase. Bit depth is the number of steps in the staircase. You can have a staircase that is 20 feet from bottom to top. One could have 20 steps, one could have 40 steps. The distance (dynamic range) is the same. More steps doesn’t alter the height of the staircase.
That is a good analogy that Bruce used to use, but it does not apply to linear files where the step size is very large in the shadows and small in the highlights as shown in Figure 2 (http://www.anyhere.com/gward/hdrenc/hdr_encodings.html) of this reference and DR is limited by bit depth; gamma encoding improves the DR somewhat for a given bit depth. These limitations are shown in Table 2. For true HDR you need log or floating point encoding. I think that Bruce's analogy would apply with these encodings.
Also, please review my references in my prior post. There is not really much room for reasonable argument.
Norman Koren (http://www.normankoren.com/digital_tonality.html)
Cambridge in Color (http://www.cambridgeincolour.com/tutorials/dynamic-range.htm)
-
Over my head but from a "users" perspective, am I better off exporting 16 bit TIFF or DNG from Phocus and then to import into LR or PS?
Unless you are doing image adjustments that require demosiacing (such as lens correction) you would be better off outputting a DNG and dealing with the raw processing in Lightroom if that is where you want to do your raw processing and also assuming that Phocus exports a raw DNG as apposed to a Linear DNG that has had the demosiacing done.
A "Linear DNG" is basically like a TIFF file in Pro Photo RGB color but with a linear gamma. DxO is an example of a raw processor that is capable our doing substantial lens based correction but needs to do the demosiacing so all it can output is a linear DNG.
So, does Phocus output a raw DNG or a linear DNG? (and do you need t do lens corrections?)
-
As a p65 user who prefers LR to C1 in some areas, I have resorted to using C1 to export DNG's that i can then work with in LR.
In light of this thread, are C1 DNGs raw or linear? I had simply assumed they are raw... until i read this thread!
And... is there an established 'optimal' workflow for those of us with raw files that LR won't work with, but who'd still much rather process images in LR than C1?
And... will LR ever be able to deal with P65 files (please)?
Thanks for any help,
Scott
-
In light of this thread, are C1 DNGs raw or linear? I had simply assumed they are raw... until i read this thread!
C1 writes out regular raw DNGs not linear. However if you want to use C1 for lens corrections (like on my Phase One 645 lenses) then you must actually process out as a TIFF.
In terms of workflow...when shooting to a card, I let C1 import and then to a DNG process output then import into LR. If I'm shooting tethered I do the same thing...
As for LR directly supporting P65+ files, yes...LR will be able to directly read P65+ files...don't know if it'll be in a 2.x update of a 3.0 version.
-
Unless you are doing image adjustments that require demosiacing (such as lens correction) you would be better off outputting a DNG and dealing with the raw processing in Lightroom if that is where you want to do your raw processing and also assuming that Phocus exports a raw DNG as apposed to a Linear DNG that has had the demosiacing done.
A "Linear DNG" is basically like a TIFF file in Pro Photo RGB color but with a linear gamma. DxO is an example of a raw processor that is capable our doing substantial lens based correction but needs to do the demosiacing so all it can output is a linear DNG.
So, does Phocus output a raw DNG or a linear DNG? (and do you need t do lens corrections?)
I found a year old thread on LL that addressed some of these issues w/o hard answers. http://luminous-landscape.com/forum/index....showtopic=31079 (http://luminous-landscape.com/forum/index.php?showtopic=31079)
I will try the analogous C1>DNG> LR and see how that works. Besides lens correction the big difference in card download > Phocus vs. card >LR are much more accurate color "out of the box" and a very big difference in shadow recovery and noise. I am not certain how much of these differences are intrinsic to the software and how much are user correctable.
Steve
-
As for LR directly supporting P65+ files, yes...LR will be able to directly read P65+ files...don't know if it'll be in a 2.x update of a 3.0 version.
Thank you Jeff.
Very exciting about LR and P65+ files.
Do you know if LR will be able to handle the 15mb Sensor+ files as well as the normal 'huge' files?
Scott