Changing Metadata resulting in lost image quality for JPEG?

[GeorgD]

(all done with 1.90beta2)

a) When I edit IPTC metadata of a JPG and save that JPG, is the image data re-encoded, so causing losses? I've the impression, as a binary before-after-comparision showed that the vast majority of a JPEG content changed...

b) When performing edit => metadata => EXIF thumbnail rotate left, only a very small part of the image changed. Hence, I guess the image data itself is not newly encoded.

c) When changing the EXIF creation date (in EXIF tab inside preview pane), only those few characters changed, hence the image data is not altered.

Did I miss any metadata changes?

@Pierre:
1) could you please verity/falsity my assumptions? You know what your code does
2) Is it possible to change IPTC edits such that no image quality loss is caused? Then, the old and limited descript.ion could be dumped for a more structured type of information storage

/Georg

[xnview]

(all done with 1.90beta2)

a) When I edit IPTC metadata of a JPG and save that JPG, is the image data re-encoded, so causing losses? I've the impression, as a binary before-after-comparision showed that the vast majority of a JPEG content changed...

b) When performing edit => metadata => EXIF thumbnail rotate left, only a very small part of the image changed. Hence, I guess the image data itself is not newly encoded.

c) When changing the EXIF creation date (in EXIF tab inside preview pane), only those few characters changed, hence the image data is not altered.

Did I miss any metadata changes?

@Pierre:
1) could you please verity/falsity my assumptions? You know what your code does
2) Is it possible to change IPTC edits such that no image quality loss is caused? Then, the old and limited descript.ion could be dumped for a more structured type of information storage

/Georg

Please post in 1.90 topic
If you edit IPTC in browser mode, it's always lossless, not re-encoding. If you are in view mode, you can use write or apply. Write make a lossless change, apply set new iptc for picture, but you must save it after, so re-encoding...

[GeorgD]

Please post in 1.90 topic

Sorry, I thought it's the same for older versions as well. Thanks for moving it

If you edit IPTC in browser mode, it's always lossless, not re-encoding. If you are in view mode, you can use write or apply. Write make a lossless change, apply set new iptc for picture, but you must save it after, so re-encoding...

Thanks for the info In browser mode, the big image stream looks different in hex viewer - can the same encoding have different hex values?

Could you rename "write" to "write lossless" or "Extend file by IPTC" or the like and "apply" to "apply to unsaved image" or the like? At least I can't remember this fine difference for a longer time, hence such a tiny hint would be greatly apprecieated

[GeorgD]

I did the test again with v1.96.5 and a more advanced binary compare method. Results: Now, XMP data is written together with IPTC. I did 3 Tests with identical changes to ITPC data.
Test 1: Changes in browser mode. A litte part (metadata area (RDF)) changed, vast majority is binary identical.
Test 2: Changes in view mode, "write" button. Result is extremely similar to test 1, but a few bites differ even though same IPTC data was entered in dialog - maybe a time stamp? Length would fit for that.
Test 3: Changes in view mode, "apply" button. Nearly everything is returned as being different => looks like re-encoding.

=> information above seems valid.

Pierre, could you please rename the buttons in view mode? I.e. "write" to "write lossless" or the like and "apply" to "apply to unsaved image" or the like? At least I can't remember this fine difference for a longer time, hence such a tiny hint would be greatly appreciated

[GeorgD]

In 1.97 still the same. Pierre, is renaming the buttons a big issue and thus has to be postponed?