[Digikam-users] re JPEG lossiness, PNG

[Remco Viëtor]

On Saturday 14 January 2012 07:44:17 Dr. Martin Senftleben wrote:
> Hi all,
> 
> I find it interesting that this has become a thread about which format
> is the best. It is a question that still bugs me, so thank you for all
> your comments! I'm still searching for the best way, but I understand
> the best would be to work with the raw images and use them as the
> initials. Right. I was always a bit reluctant to work on raws, because
> it seemed quite complex, but it doesn't seem to be with digikam.
> I took one raw image now and saved it in different formats, without
> any changes applied.
> The outcome is as follows:
> cr2 (original from camera of course): 23.7 MB
> jpg (produced by camera): 5.7 MB
> jpg (produced by digikam from cr2): 1.6 MB
> tif: 51.7 MB (w/o compression, w/ compression it's 27.4 MB)
> pgf: 22.8 MB
> png: 22.3 MB
> jpg2000: 22 MB
> 
> pgf is for me out of the question, I had too much trouble with that
> format. jpg is lossy, tif quite big (even with compression). Remain
> png and jpg2000, where jpg2000 is a format not so well supported.
> 

Personally, I save and tag my raw files, then convert to (16-bit) PNG for 
editing, and use JPG for the final images, after resizing to the required 
size.

> What puzzles me (is this a bug?) is that none of the tags which I had
> saved with the raw file have been taken over by any of the exported
> formats. It would be very difficult to create new images from the raw
> file and again and again add the tags to the new versions. Only the
> description and other exif data were taken over - at least something.

I saw that as well, in that Digikam showed tags with the converted image, but 
tags and description were not stored in the image file itself. (and I have 
selected all but the last three options in the 'Settings/Metadata/Common 
metadata actions' section). To get that information in my png files, I used 
the menu-item 'Image/write metadata to image'. But I'm not sure if that works 
on a series of selected files.

> What workflow would be the best to make sure that I have not to repeat
> adding tags to the different versions of the same file? Or is there
> some setting that I missed which would do the trick?
> I could think of: loading raws from the camera, then creating copies
> and then starting to tag them etc. But this would only be ok for the
> first version I create. As soon as I start to create another version
> with some other settings from the raw file, I'll have to redo the
> tagging... 

Do you use XMP files? I get the impression that the interactions between in-
file metadata and XMP metadata isn't quite optimal yet.

> Another limitation, if I use png as the format to work
> with, is that the files are too big for uploading to flickr. This will
> always add the step of converting to jpg to the workflow. It's not too
> much work, but it's a step that requires some attention and time,
> depending on the number of images I want to upload. What if
> flickrexport would do the conversion?

For that kind of things I use batch mode (with the tools 'resize', 'sharpen', 
'add metadata template', 'export to jpg'). This also means I never resize my 
pngs (I do crop them to get the optimal aspect ration and composition).

I prefer working from raw files, as I get better results from them. But it 
depends also on your kind of photography: if you shoot events yielding 100s of 
photos to be processed quickly, you might be better off using in-camera JPG 
(or save as both JPG and RAW to use the raw for the best shots). If you're 
mostly doing subjects where you end up with a few (10s) of images that are 
kept/processed (macro, portrait, landscape), RAW might be the preferred 
starting format, for the control and quality it can give you.

Just a few points if you start working in RAW.
Try to do as much as possible in the RAW converter: white balance, exposure 
correction, noise reduction if needed, etc. This can reduce artifacts and 
information loss, as some steps can be done before demosaicing. Though I 
basically ignore the Post Processing tab, as changes there are done after 
demosaicing, and the normal edit page gives me more control.

And, don't forget to sharpen your images. At least apply an unsharp mask at 
the end of your edits (for a 10MP pic, radius 25/amount 0.6/threshold 0.03 
seems to work nicely for me as a starting point, you'll have to adapt it to 
your image). Just don't overdo the sharpening to avoid artifacts.
(If I really want to pull out all the stops, I sharpen 2 or 3 times: 
- capture sharpening  after raw conversion: refocus on default settings
- unsharp mask after edits as mentioned above
- unsharp mask after resizing for final output; usually radius=10, amount to 
taste)

Regards

Remco