Processing 16 bit images with layers: only option Photoshop?

[Jekyll]

I am just trying to get the hang of the post processing my digital photos. Up to now my main focus (no pun intended has been to get best shots I can in camera (on my 5D mk III). I am following the frequent advice to shoot raw. But now I am needing to do a little digital editing to prepare some photos for sharing (e.g. flickr for athletes I coach or my web page for business) and printing (on my own high quality printer - also 16bit - on professional 'glossies' and in photo books I want to produce).
My software to do this is limited to software that came with my Canon and PSE 10 but might work for now, although I will invest in new if/when I need to. My first problem is around bit depth. The raw photos on the 5D have 14 bit colour depth according to spec sheet. I can maintain this high quality through ACR if I use 16 bit option, but if I want to do any pixel level editing (e.g. red eye) it seems the only software that let's you use layers in a 16 bit image is full version of photoshop. Other options like PSE and GIMP only process layers in 8 bit/channel images.
Am I missing something? Any suggestions or do I just need to bite the bullet for full version of PS6? I have read other posts about pros of using LR4 and I may start there. But I don't to go down a path that blocks me from highest quality output. I understand that for flickr and web pages 16 bit/channel is probably overkill, but for my own prints the 16 bit prints seem to be better.

[Alandb]

I may be asking the wrong question but are you sure you want to use layers? If you have already invested time an energy into PSE 10 and are comfortable with that then, yes, you might want to invest in PS6. If not, you might want to consider LR4 and, instead of layers, look at the NIK suite of plugins. I looked at PSE 9 and layers and found it cumbersome. On the other hand, I felt much more comfortable with the Control Points concept of the NIK system. You can download trial versions of almost any software. LR 4 gives you the additional plus of file management.

[Graystar]

First, the 14-bits of the camera and the 8/16 bits of the image have nothing do to with each other so don't worry about that.

Second, Lightroom doesn't have layers.

Third, your display system is an 8-bit system...no matter how you process the image, it's always transformed into an 8-bit image for display and printing as well unless you're using a 16-bit printer driver.

The place where 8/16 bits matters is when you're trying to increase the difference in luminance between two very similar colors. Think of a compressed accordion. The bellows has a fixed number of folds in it. As you stretch out the accordion, the bellows is much longer, but the number of folds is still the same. This is the problem. If there aren't enough values between two colors, and you try to "stretch" them out, you'll get banding. The actual occurrence of this is very rare, and is usually only a noticeable problem with very smooth gradients in B&W images. Usually, adding a small amount of noise fixes the problem. In any case, this is not a layer problem. You can get a lot done with 8-bit layers if you're not working in B&W (and even if you work in B&W, it's rare when you'll have this problem.)

Note that the PS layer system does have greater functionality than PSE. Whatever process you're interested in, make sure that the functions being used are available in PSE. You may need PS just to be able to apply a particular adjustment layer or to make adjustments to a luminance channel.

And don't forget...even if it's not a problem in the image data, you may perceive a banding problem because of your 8-bit display (The solution to this particular problem is to change the info window to 16-bit and review the actual values of the pixels to see if the banding is real or created by the display transformation.)

Finally, understand that 16-bit data, in of itself, doesn't give an image more "quality" than 8-bit data. 16-bit data avoids certain issues within a fairly limited set of image-manipulation operations. Otherwise, 8-bit operations will typically produce results that are visually indistinguishable from 16-bit operations.

[jcuknz]

If you don't want to spend on PS then you could try Paint Shop Pro which is at a comparable price to Elements and of course has layers since way back. I would hate to be without layers .... urrrgh .... a horrible thought

Graystar Finally, understand that 16-bit data, in of itself, doesn't give an image more "quality" than 8-bit data. 16-bit data avoids certain issues within a fairly limited set of image-manipulation operations. Otherwise, 8-bit operations will typically produce results that are visually indistinguishable from 16-bit operations
Thankyou for that reassuring comment becuase I am not sure about this 'bit' business.

Edit ... I have never sighted GIMP but I imagine that is another alternative to PS to have layers.

[revi]

Yes, GIMP also has a layer system, even though it's a bit different in some aspects from PS's (no adjustment layers). The current version is still
8-bit, although internally it can work in higher precision (discussed in another thread).

But, iirc, PS is not completely 16-bit yet, either. Some tools and filters are still 8-bit only (as are some plug-ins, I suppose). Not sure how important
that is in practice (if you want to see how 'disastrous' 8-bit editing is for your photos, look at the work of Donald (McKenzie), a GIMP user )

[Jekyll]

First, the 14-bits of the camera and the 8/16 bits of the image have nothing do to with each other so don't worry about that.

Second, Lightroom doesn't have layers.

Third, your display system is an 8-bit system...no matter how you process the image, it's always transformed into an 8-bit image for display and printing as well unless you're using a 16-bit printer driver.
.

Hi Graystar,

Thanks for your comments. On the 14 bits in the camera part, I did some digging and I think you may be mistaken. There is an excellent explanation from Canon of bit depth in camera and in images at http://cpn.canon-europe.com/content/...chnical/raw.do and this explains how the raw images from a Canon 5D mk III are indeed 14 bits per channel raw images. These are reduced to 8 bit per channel for jpegs produced in camera, but the full 14 bit per channel is maintained in the raw image - which is why you can avoid some of the artefacts that describe.

I understand that Lightroom doesn't have layers, I was thinking about using it for cataloguing and some other functions as well as raw processing.

And yes, I am using a 16 bit printer drive on my mac for a Pixma Pro-10 printer.

[Graystar]

Hi Graystar,

Thanks for your comments. On the 14 bits in the camera part, I did some digging and I think you may be mistaken. There is an excellent explanation from Canon of bit depth in camera and in images at http://cpn.canon-europe.com/content/...chnical/raw.do and this explains how the raw images from a Canon 5D mk III are indeed 14 bits per channel raw images. These are reduced to 8 bit per channel for jpegs produced in camera, but the full 14 bit per channel is maintained in the raw image - which is why you can avoid some of the artefacts that describe.

I understand that Lightroom doesn't have layers, I was thinking about using it for cataloguing and some other functions as well as raw processing.

And yes, I am using a 16 bit printer drive on my mac for a Pixma Pro-10 printer.

That's some nice marketing material, but it's misleading.

First, my background on this. I'm a computer programmer by trade, and I've reviewed the source code of DCRAW, an open-source RAW processing library. I've also reviewed the algorithms for transforming RAW data in to image data, and how colors are processed in images.

That Canon page is comparing RAW output to an out-of-camera JPEG. The way JPEG compression works is to take areas with few levels between the colors, and turn them into one color. As I said previously, and as also implicated in the page you linked to, banding can occur when trying to change the luminance in areas where colors are very similar...

"The banding is a result of the few levels in the shadows being stretched by the exposure compensation"
(page 2 of linked page)

So of course, if you take an image that has already experienced lossy compression, zoom into an area where the data was "smoothed", and try to expand the luminance, you'll get banding. Big surprise.

But we're not talking about JPEGs. We're talking about 16-bit RAW conversions that must be converted to 8-bit for particular operations. A more appropriate comparison would start both images from RAW, convert one into 16-bit and one into 8-bit, and then process the images from there. If they did that there would be a drastically different result.

The article also makes an error when it states that the 12/14 bits of RAW data are reduced to 8 bits. That's not what happens at all. RAW data and the data in a TIFF or JPEG represent different information that's not compatible. RAW data represents, directly, the luminance levels that were recorded at individual photosites on the sensor. Image data, on the other hand, are pointers into the color space of the image.

For example, the RAW value 512 may turn into the RGB value of 100 for an image in the sRGB color space, or the value 80 for a image in the ProPhoto color space. It's the same color in both images...but the image data is different because the color spaces are different. The image data alone isn't enough to describe the color. You need a color space to fill in the blanks. That's why image data is not the same as RAW luminance data.

So to say that bits were reduced is simply wrong. What happens is that the RAW data is demosaiced in the camera's color space. Then that image is transformed into the selected color space. The image is transformed by first mapping the demosaiced colors to a Profile Connection Space. The PCS is a very large color space that has all colors that a human can see. Then the process is reversed where the colors are mapped from the PCS to the selected color space (sRGB, Adobe RGB, ProPhoto, whatever.) If there are colors in the PCS that don't fit into the selected color space (sRGB is a fairly small color space) then the color is process in a particular way, which I won't get into. The point is, that when you're done with this process there's really no DIRECT relationship between the RAW values you started with and the final RGB values in your image. It's an indirect relationship through the color space transformation process.

The final point I'd like to address is that of color accuracy. Some people might say that 16-bit is more accurate that 8-bit. Well, the starting values that are created from RAW are basically guesses. Every pixel on a sensor records only one color...the other two colors are invented from the surrounding pixels. Their values are based on whatever demosaicing algorithm you decide to use (such as ACR, eahd, hphd, amaze, vng4, and others.) In fact, even the recorded color is adjusted by the demosaicing algorithm. So this idea of color accuracy isn't so straight-forward. Also, when viewing an image it's impossible to see a change of 1 in one color channel of 8-bit color (say, from 100 to 101.) What we're talking about with 16-bit is 256 choices between 100 and 101. These are differences that simply cannot be perceived.

As I said, 16-bit is valuable in two instances...smooth gradients in B&W images, and a very narrow set of luminance operations on very specific data structures. Otherwise, it's pretty much an unnecessary level of detail.

And yes...I shoot RAW and do everything in 16-bit.

[revi]

Graystar: would you have a reference explaining the use of a PCS in RAW development?

I'm a bit surprised that the conversion would go through such a large colour space (after demosaicing), as such a step
in itself could give rise to banding (in theory at least): a large colourspace implies a large difference between neighbouring
values, where a smaller space could place several colours. So transforming an image with gradients of fairly unsaturated
colours would throw away quite a bit of information.

As far as colour accuracy is concerned: the initial image after raw development will be as accurate in 8 bit as in 16 bit (as long as
display and printing is predominantly done in 8-bit accuracy). The difference is in what happens during further processing...

[Graystar]

Graystar: would you have a reference explaining the use of a PCS in RAW development?

I'm a bit surprised that the conversion would go through such a large colour space (after demosaicing), as such a step
in itself could give rise to banding (in theory at least): a large colourspace implies a large difference between neighbouring
values, where a smaller space could place several colours. So transforming an image with gradients of fairly unsaturated
colours would throw away quite a bit of information.

As far as colour accuracy is concerned: the initial image after raw development will be as accurate in 8 bit as in 16 bit (as long as
display and printing is predominantly done in 8-bit accuracy). The difference is in what happens during further processing...

Unfortunately I don't have a good reference explaining this stuff.

If you're a programmer you can review the DCRAW code found here...
http://www.cybercom.net/~dcoffin/dcraw/

Here's a nice reference on demosaicing...
http://www.unc.edu/~rjean/demosaicing/demosaicing.pdf

And of course, Doug Kerr can be relied upon to write a detailed description of just about anything...
http://www.openphotographyforums.com...ad.php?t=11262

The color space transformation is nothing special. That's how all color space transformations occur. The PCS is always either CIELAB or CIEXYZ. There's no risk of banding because nothing is being expanded. The color points from one space are simply located in the PCS, and then that color point is used to convert to the target space. The CIE spaces are floating point spaces...not limited to 16-bits (actually all spaces are floating point...that's why you can have images with 24 or 32 bits per channel. The limit is in our storage mechanism. For example, JPEG explicitly states 8-bit...but it could be 8-bit ProPhoto.)

From the Adobe RGB specification

4.3.1.3 Color Space Encodings
The value range for Adobe RGB (1998) color space component values shall be [0, 1].
The color component values shall be encoded using integer or floating-point encodings.
Integer encodings shall be unsigned with 8 or 16 bits per component with the same number of bits
for all three components. The R’, G’, B’ component value range [0, 1] shall be encoded over the code
value range [0, max integer value]. R’, G’, B’ code values of 0, 0, 0 shall represent the color space
black point, and max integer code values shall represent the color space white point.
For integer encodings, all code values shall be within the color space gamut.
Floating-point encodings shall be 32 bit per component using the floating point encoding defined for
the applicable image format. If no such encoding format is defined, then use IEEE 754-1985. In
floating-point encodings, a component value and its encoding value are the same. Code values 0.0,
0.0, 0.0 shall represent the color space black point, and code values 1.0, 1.0, 1.0 shall represent the
color space white point. Component values outside the range [0, 1] are not allowed for floating-point
encodings.

If you look in the spec you'll find sections such as "4.3.7.3 Converting to ICC CIEXYZ PCS values". That's basically what a color profile is...it contains the color space definition and methods for converting to and from the PCS. It's the only way to transform the color space of devices such as cameras and printers into the standard color spaces.

http://www.adobe.com/digitalimag/pdfs/AdobeRGB1998.pdf

[pnodrog]

First, the 14-bits of the camera and the 8/16 bits of the image have nothing do to with each other so don't worry about that.

.

You said that Canon had nice marketing material but it is misleading!

To say the 14 bits of the camera and the 8/16 bits of the image have nothing to do with each other is also misleading. From that statement the conclusion I could make is that I do not need a camera.

Simplification is fine - over simplification is misleading. Maybe you should revise or clarify the statement. Even if the relationship is indirect it still exists.

[Graystar]

You said that Canon had nice marketing material but it is misleading!

To say the 14 bits of the camera and the 8/16 bits of the image have nothing to do with each other is also misleading. From that statement the conclusion I could make is that I do not need a camera.

Simplification is fine - over simplification is misleading. Maybe you should revise or clarify the statement. Even if the relationship is indirect it still exists.

That statement is accompanied by a detailed explanation that provides all the clarity necessary to understand the difference. Nothing more is needed.

[GrahamTebby]

I am just trying to get the hang of the post processing my digital photos. Up to now my main focus (no pun intended has been to get best shots I can in camera (on my 5D mk III). I am following the frequent advice to shoot raw. But now I am needing to do a little digital editing to prepare some photos for sharing (e.g. flickr for athletes I coach or my web page for business) and printing (on my own high quality printer - also 16bit - on professional 'glossies' and in photo books I want to produce).
My software to do this is limited to software that came with my Canon and PSE 10 but might work for now, although I will invest in new if/when I need to. My first problem is around bit depth. The raw photos on the 5D have 14 bit colour depth according to spec sheet. I can maintain this high quality through ACR if I use 16 bit option, but if I want to do any pixel level editing (e.g. red eye) it seems the only software that let's you use layers in a 16 bit image is full version of photoshop. Other options like PSE and GIMP only process layers in 8 bit/channel images.
Am I missing something? Any suggestions or do I just need to bite the bullet for full version of PS6? I have read other posts about pros of using LR4 and I may start there. But I don't to go down a path that blocks me from highest quality output. I understand that for flickr and web pages 16 bit/channel is probably overkill, but for my own prints the 16 bit prints seem to be better.

[GrahamTebby]

Try http://www.sagelighteditor.com/ - I think it excellent.

[arronlee]

Hi, Jekyll.
I think your problem is not so complicated to solve, and Photoshop is not your only choice, I want to say. There are so many tools online dealing with image processing. You can choose one whose way of processing is simple and fast. It can save a lot of time for you. You can do any pixel level editing (e.g. red eye) with 16 bit option with the help of it. I hope you success. Good luck.