Gene51 Loc: Yonkers, NY, now in LSD (LowerSlowerDelaware)

Read the article - and it is very convincing. It flies in the face of convention, though. I do like the better image quality, especially in extreme situations, that 12 bit offers over 14 bit. At least with the D810. I suspect that each camera handles dynamic range and headroom a little differently.

BHC Loc: Strawberry Valley, JF, USA

My understanding is that 12 bit yields 4096 colors while 14 bit yields 16384 colors. I guess it depends on how sensitive your computer, eyes and printer are to color variation.

JimH123 Loc: Morgan Hill, CA

A more accurate way of saying this is that the three colors, red, green and blue, can each produce 4096 separate levels for 12-bit, or 16384 separate levels for 14-bit. You could say that 12-bit could produce 4096 x 4096 x4096 colors. Or 14-bit could produce 16384 x 16384 x 16384 colors.

A definition of levels is really thought of as 4096 red levels and 4096 green levels and 4096 blue levels for 12-bit.

It is faily easy to see the 8-bit limitation with JPG files. If you do a curves adjust on a JPG image, and there is some clear blue sky, you can generate banding on the sky. For instance, there are 256 separate blue levels. When you create an S-curve, most of those levels are used in the middle where it is more linear, and very few levels are available at the ends.

BHC Loc: Strawberry Valley, JF, USA

In other words, 12 bit is approximately 68.7 billion colors and 14 bit is about 4.4 trillion colors.

My eyes and equipment are definitely not that discerning!

Thank you for the correction.

JimH123 Loc: Morgan Hill, CA

Yes, that is correct. I've seen crayon boxes with lots of different colored crayons, but not that many colors!

And you are correct. You are not going to resolve this with your eye.

I know I can stretch a JPG image with the curves adjust so that I can see banding. But I do not think I can do this with 12-bit. Hopefully, everyone knows what I mean when I describe the curves adjust. Photoshop has it. PSE leaves it out.

I have pasted in an example I found doing a Google search for banding.

I might add that the difference from 8-bit JPG to 12-bit RAW is 4-bits. Those 4 bits would divide each of those bands in the image into 16 separate bands, which though small, would not be invisible. To go to 14-bit is 2 more bits, so instead of dividing each band into 16, it would divide each band into 64 bands. And I know I could not resolve that.

sample banding image found by googling

(Download)

Apaflo Loc: Anchorage, Alaska

Not quite right yet! It gets worse... Because those calculations are based on the idea of 3 different color channels, as is true with a TIFF or a JPEG. A 12-bit or 14-bit TIFF file would indeed have that many possible colors. But the raw sensor data has color encoded using a Bayer Color Filter array.

Color in the raw file is a function of a matrix of sensor locations. Typically the minimum size of a matrix is 4x4, but the average pixel in an image is demosiaced using a 5x5 or even an 6x6 matrix. The number of colors that can be decoded from 14-bit raw data using the minimum of a 4x4 matrix is 2 ^ (4 * 4 * 14). That is as opposed to an 8 bit JPEG RGB image that is 2 ^ (3 * 8), or a 16 bit TIFF that is 2 ^(3 * 16).
The actual difference of greatest significance is not the number of colors at all, but the number of brightness levels (regardless of the color). The raw sensor data uses linear encoded data, so a 14 bit file can have a maximum 16384 brightness values over 14 fstop range. But JPEG and TIFF files are gamma corrected, so an 8 bit file does not have 256 values, only about 240. But those 240 values are spread over almost a 9 fstop range. A 16 bit gamma corrected TIFF file could encode only something like 61,000 values, and over an 18 fstop range except that the data comes from a raw file that cannot produce that many. (With a 16-bit RAW file that is possible, but no DSLR is currently using a 16 bit ADC.)

The reason to use a 14 bit RAW format is the added dynamic range. The advantage is subtle, and cannot be seen in every image. It also is not necessarily of great significance with some cameras. As technology improves the higher bit depth is more important, and we can expect that soon enough DSLR's will have 16 bit RAW file formats.

BHC Loc: Strawberry Valley, JF, USA

per http://photographylife.com/14-bit-vs-12-bit-raw:

(Download)

Apaflo Loc: Anchorage, Alaska

That is wonderful, and it actually is true that an 8 bit JPEG can show 16.8 million different colors, and it is true that that is plenty enough.

But the article entirely misses the point that it isn't the number of colors that makes a difference between 12-bit and 14-bit raw sensor data. What does make a difference is the maximum dynamic range the data can encode. Linear encoded data can hold:

Dynamic_Range = 6.02 * N +1.76 dB

Where N is the number of bits. Hence a 12-bit encoded RAW file can have a maximum dynamic range of 74 dB and a 14-bit encoded file can have a dynamic range of 86 dB, or essentially 12 fstops for 12-bit files and 14 fstops for 14-bit files.

The reason HDR requires multiple exposures is that nature commonly lights up scenes with more than 20 fstops of dynamic range. Even with a 14-bit raw file we can't come close to recording that with a single exposure.

What we really want is a camera that uses 24-bit RAW files!

Hammer Loc: London UK

Hi,

Very interesting and many thanks . Now you have explained why these files are called 12 and 14 bit . Why wasn't I inquisitive enough to question this . Now back to saving 14bit but with lossless compressed.

Starting to understand .

Gene51 Loc: Yonkers, NY, now in LSD (LowerSlowerDelaware)

Jim and Mogul, here is are some other views that supports John Sherman's experiments. :

http://www.dpbestflow.org/camera/sensor#depth

http://photo.stackexchange.com/questions/17616/is-14-bit-raw-better-than-12-bit-raw

http://www.earthboundlight.com/phototips/14-bit-raw-12-bit-part-two.html

The writers at ASMP, DXO, Earthbound Light, John Sherman, etc all seem to have to have come up with the same conclusion - though there is a clear and demonstrable mathematical difference, it exists on paper, and in practical use human vision lacks the "bit depth" to be able to see a difference - even when processing images that have been severely underexposed. When you factor graphics bit depth, print gamut and bit depth, the average person, and even a critical photographer, will have a hard time seeing the difference. Let the mathematicians argue - I would not worry much about this. If your camera performs/functions better at 12 bit uncompressed than at 14 bit, I see no practical reason to shoot 14 bit. And there are quite a few articles written and comparisons illustrated with practical images that seem to support that notion.

woolpac Loc: Sydney Australia

I think dynamic range is an expression of the ratio between d-max/d-min on a linear curve before compression in other words white with no detail to black with no detail and generally expressed in stops.

Bit, be it 12 or 14 is a description of color depth as previously mentioned.

selmslie Loc: Fernandina Beach, FL, USA

The first two are correct but not the last one. The raw file contains only 2^14 brightness values (not colors) at each pixel and these get combined or averaged with the neighboring pixels to generate either a n 8-bit or 16-bit color JPEG or TIFF. The maximum number of color permutations is mathematically limited to 2^24 or 2^48 values since any in-between values resulting from the combining or averaging of raw pixel values are either rounded or truncated.

But these are theoretical distinctions. An sRGB or Adobe RGB gamut does not contain all of these values. Particularly, the darkest red, blue and green combinations are rendered simply as black.


Not correct. It has all 256 values, it's just hard to see the extreme values.


Yes and no. The added dynamic range is not present in every scene - some have less and some more than a 14-bit raw file can deal with. That's why we need HDR at times.

What you see in the image, on a print or on your screen, is a fixed dynamic range from Dmin to Dmax, depending on your medium, not the original dynamic range of the scene.

lamiaceae Loc: San Luis Obispo County, CA

Now I have a question (for a change). I did not know you could select 12-bit vs. 14-bit vs. 24-bit in your camera. I thought that was a fixed property of the processor in the camera. I am familiar with changing bit depth in PP in ARC and Ps say from 8-bit to 16-bit to 32-bit (huge) files and various resolutions, 72ppi, 300ppi, 600ppi,etc.

My understanding, I am mistaken, is that older early digital cameras were 10-bit, than 12-bit, and that most today are 14-bit with few high end cameras being 24-bit. And I am referring to bits not megapixels.

Yes, I shoot RAW >95%.

selmslie Loc: Fernandina Beach, FL, USA

Not all cameras offer a choice of 12- or 14-bit. This is the raw bit depth.

When we refer to 24- and 48-bit files we are normally referring to JPEG and TIFF color image files. There are other possible combinations.

Apaflo Loc: Anchorage, Alaska

Those two cites are excellent examples of how you can find virtually any opinion you'd like to on the Internet, if you aren't critical.

The second one has so many misconceptions it is safe to say none of the participants has any clue about digital data. The first one isn't as bad, but says this, "First we should dispel the myth that higher bit depth translates into higher dynamic range. It does not." Opps, write that guy off too! Clueless about digital data encoding, and therefore nothing he says on the topic is worth listening to. (See the formula given in a previous article about the relationship between dynamic range and bit depth. Any book on practical systems engineering will derive the formula and explain the significance to audio, video, and image data.)


This cite is different. It actually makes sense, and the author understands what he is talking about! I guess you assumed he was saying something other than what he did, because he doesn't agree with the others and says, the differences "are real enough to further warrant shooting in 14-bit raw".


You really need to read for detail. They say that is true, if what you want to see is a difference in color. There is none.

There is a distinct difference in dynamic range, and while that obviously does not affect all images, it does affect some very significantly.


No they don't! For example, how many cameras can even shoot "12 bit uncompressed"? None of the current Nikon APS-C cameras shoot uncompressed. The D7xxx series are the only ones that can shoot lossless. For full frame Nikon cameras I'm not sure which ones can shoot uncompressed, but the high end models certainly can while the low end (D610) cannot.