Work flow and data loss.

[Manfred M]

Ted - I think your assessment is bang on.

Your typical modern camera (Sigmas aside?) tend to capture a good 12 or more stops of dynamic range and can encode data to 12 to 14 bits. We even have to fit them with UV and IR filters to ensure that they only capture the "visible" range of light. In other words they can significantly outperform any of the devices that we view the output with, The best commonly available screens approach being able to produce 10-bit colour (their gamut being AdobeRGB compliant) and I suspect a dynamic range that might be around 9 or 10 stops (I'm not sure if you have better data, based on your research).

Printers (depending on the specific printer being used) can have a slightly larger colour space than AdobeRGB, but the dynamic range of a print is probably in the order of 5 or 6 stops.

So in theory, best case out there, our modern cameras can likely capture most of the colours visible to humans; i.e. the reference CIE 1931 colour space that defines what an "average" person can see. To me, I usually care less about the total millions and billions of colours that the various devices can reproduced, but rather about the boundary conditions that define the edge of what can be reproduced. Based on the what I've been able to determine, the highest number I've seen suggests that best case is that humans can see around 10 million individual shades, not the billions of shades that can be recorded and reproduced by the equipment we use.

So in the scheme of things we collect a lot more data than we will ever use, in most cases. In casual editing, changes are we will never use all the data available to us. If we totally blow our exposure, mess up the white balance or use some fairly sophisticated techniques to recover shadow and highlight details in post-processing we might want to dip into this "surplus" of data in the raw files.

Where we can get into trouble with 8-bit data is when we make extreme edits that results in the editing software merging different colours with discrete, visible steps. This can show up anywhere, but banding in the sky is where we often notice it. Editing in 16-bit can definitely be one way of solving this issue,. I feel that Ted is quite right, if we do those extreme edits in our raw editor, we can avoid this issue and do the heavy lifting in 16-bit; then it really doesn't matter whether we use 8-bit or 16-bit in the pixel based editor.

The only place I have found a clear advantage to working in a wider colour space using a raw file as a starting point is when using a professional photo printer (one which has icc profiles published for it by the paper manufacturers), especially if there are brilliant colours in the image. I've always used a 16-bit workflow here so can't comment on how well 8-bit will work.

[JBW]

I' To me, the usage of "loss of information" or "loss of data" has varied confusingly in this thread. Yes, here I go again with a usage mention. So . . .

. . . suppose someone here produces a perfect horizontal gradient 1024px wide by say 50px deep, using the incredibly capable Photoshop (PS), or even four gradients, namely grayscale, red, green and blue. Suppose now we save as 16-bit TIFF and then again but as 8-bit TIFF.

Now suppose we bring the TIFFs up on our output device of choice. Will we see "vertical banding" in the 8-bit version(s) at 100% and at a proper viewing distance . . . . I think not.

Therefore "information" i.e. what we actually see in the form of visible lightness or visible color information, has not been lost - although "data" certainly has because 256 tonal levels got compressed into 1.

If so, then the root fear here must be of the potential loss of visible data. We all know the causes of that and it's various appearances but I claim that they are not caused per se by the simple conversion from 16-bit depth to 8-bit depth.

So, one could convert a raw file directly into an 8-bit file, tweak it a little in GIMP pre-V2.9, e.g. crop, exposure comp, sharpen and save it quite satisfactorily as a 90% quality 4:4:4 JPEG, post it here and I'll bet that nobody would know how you got there

As anyone who has tried to talk to me about tech stuff knows i am not a techie. But what you say about 'loss of visible data' pretty much clarifies what I am trying to avoid. I agree with your assessment of who could tell because my one and only published shot was shot in JPEG and processed in Gimp.

But and it is a major 'But' of late people have been commenting on how much nicer my colours are. This improvement is due to a change in camera and a change in software. The software and the camera/lens have been crafted to work together.

It works better in both RAW and JPEG. I still go through Gimp for cloning and a few tweaks to get to posting on the web. But using software designed for the camera lens combination I use certainly lessens the potential to lose visible data.

I realize how frustrating you techies find my lack of precision in explaining this but I have confidence you can translate my thoughts into techie thought

[Manfred M]

Brian - The nicer colours you have noted are primarily due to a more modern camera with a superior lens and large sensor versus what you had in your superzoom. Post-processing can enhance you image, but the base quality of the camera's optical system are ultimately going to be the source of the improved quality you have seen.

[JBW]

Brian - The nicer colours you have noted are primarily due to a more modern camera with a superior lens and large sensor versus what you had in your superzoom. Post-processing can enhance you image, but the base quality of the camera's optical system are ultimately going to be the source of the improved quality you have seen.

I did try some of my 4200 shots in SE and you are right there was not much difference between SE and Gimp alone for the 4200