|
I recommended saving the files to Tiff and seeing if the issue remains, so far I don't think you have, instead you tried a DNG and unfortunately DNG was doomed to failure because it suffers the same issues that peg suffers.
I didn't go into any depth in my post thinking it was enough to suggest the Tiff test, and not wishing to create an over lengthy post.
I now think I need to explain more. I will keep the explanation simplistic for brevity, those with more knowledge please forgive the inaccuracies forced upon me in keeping it simple.
In a film image all light intensities are analogue and therefore there is an infinite gradation and no banding can exist. Digital attempts to mimic this within the limitations of the data files and the manner of capture.
It turns out that there aren't an infinite range of tones in a digital image or anything like it. The tonal range of Red Blue and Green are limited quite severely to just 256 light intensities in each colour at any pixel. That's it, that's all you have.
This means that the best you can hope for is striping and striations as all light intensities at all pixels are clumped into one or other these really quite minimal intensity levels. It doesn't matter too much with fine details adjacent to each other, but in broad expanses of colour with minimal light intensity changes, you will have banding present.
The reason why 256 is chosen as the entire allowable range is a computer limitation due to the 8 bits word size of computers and how digital data can be represented.
Nonetheless 256 light intensities are sufficient to prevent noticeable banding and give a smooth looking photographic appearance successfully mimicking analogue images in gradation.
So the raw data in the Raw file is sufficiently smooth for our requirements in representing the scene, and if this was far as we went things would be fine, but we also need to edit and save the image.
Two things now come into play and they are a disaster for the image.
The first is the need to post process.
As you modify the image with some processes any banding which was invisible can end up being emphasised or enhanced until its visible. For example cropping so that any banded areas are enlarged coupled with changes to the tonal range which reduce or cut the upper and lower values, could severely reduce the available intensities. Such changes are colour balance coupled with contrast changes plus a levels change.
These manipulations are severe in impact and they could conceivably reduce the available intensities from the original 256 down to as low as 110 or even less and the levels change can magnify those 110 levels to backfill into all 256 levels thus magnifying and making visible any banding present.
The changes you make in post processing can therefore have a serious effect on the image and enhance banding. This is one reason why getting it as good as you can in camera to minimise post processing is a good idea, and more importantly having a workflow that has minimum impact on image quality is paramount.
The second is the chosen output format.
The output format chosen for the image whether it be Tiff Jpeg DNG or whatever can have very serious consequences on the quality of the image.
Images especially modern images are huge. they contain enormous mounts of data to record light intensities at each pixel. lets work it out for a humble 12 megapixel image.
256+256+256=768 (256 x Red Blue Green) 768 x 12,000,000 = 9,216,000,000 seperate pieces of information needed to represent 256 light intensities in red blue and green at all locations of a 12 megapixel image. This is an enormous amount of information.
Thats why a Tiff is enormous. Only a Tiff contains all the picture information with the minimum of banding present, and that's why the very best images are saved as Tiff.
The trouble is Tiffs are huge, so jpegs are often chosen. Jpegs are tiny compared to Tiffs and much easier to store. How can this be if we need a Tiff to hold all the data.
Well the only way it can happen is if you don't hold all the data. What you do is you look at the light intensities for similar values and store them all as the same value in other words if you see blue intensities adjacent to each other in the range 111 112 113 114 115 116 117, you clump them all together and call them light intensity level 114. The result is you have magically reduced the files to a much more useable size, and in most images nobody will notice that most of the image information is now in the garbage and has been thrown away, the images look the same.
If however you have broad expanses of the same colour, you will notice banding suddenly appearing.
And that is what you are seeing in your images, enhanced banding due to the losses inherent within the Jpeg method of compression.
And it gets worse. When saving Jpeg images you are given a choice of high and low quality, choose the high quality and you can get jpeg banding and pixellation issues. choose the lower quality to have lovely small files and you are guaranteed to have the pixellation and banding issues.
Look at the file sizes. My own camera images can be outputted as a 50 megabyte Tiff or a 5 megabyte Jpeg in high resolution. The file size says it all, effectively the Jpeg has thrown away 90% of the image. When the Jpeg reconstructs this image from the remaining 10% the only outcome can be banding and pixellation. Personally I am just amazed that Jpegs look anything like the original image with the amount of processing and compression they use.
Jpeg is called a "Lossy" storage because it loses important image data. Tiff is called "Lossless" because it loses nothing of the image.
What of DNG, well DNG is also a very small file size, and the reason should be obvious, DNG is also a Lossy compression, and DNG files also introduce pixellation and banding.
I should state that Jpeg and DNG compression are very sophisticated and whats actually happening is much more sophisticated than I have suggested. but the principle is the same. Different Jpeg and DNG engines will give different results but all Lossy compression will introduce pixellation and banding.
So whats the conclusion.
The conclusions are,
1 The more you post process an image the more you will expose and enhance the image defects that digital images contain.
2 Jpeg and DNG image formats introduce banding issues into images and make them obvious.
3 If you want all the image data preserved with the highest quality and want to minimise artefacts and banding issues there is only one choice. - Tiff.
Tiff is the choice of professionals and archivists, Jpeg is the choice of everyone else.
I will add that some professionals due to other constraints, such as journalists on location save Jpeg for its quick data transfer to meet deadlines for publication. Jpeg is also the only way to distribute files, so you have to accept its limits and shortcomings.
I recommend saving some test files as Tiffs and seeing if the banding disappears.
| 
Originally posted by dcpropilot 
Post #12 by Oldbayrunner
Similar Threads 
