I've done some testing recently. When I edit or save a jpeg image (originally shot in RAW and converted to jpeg) I can see degredation of the image, as expected. However, I also tried this on some TIFF images and they also lost detail, not as much as the jpegs, but they still lost detail. I though TIFF was a "lossless" format. Why is this happening?
Honestly, I'm normally not a pixel-peeper like this, but it bothers me when I'm loosing detail in a supposedly lossless format. Am I missing something here?
Interesting situation - I doubt you're discovering that TIFF is losing information though. The second image, which you describe as "edited in TIF" is actually a jpeg file (which of course can introduce some loss). Not only that, it's been compressed pretty highly. So I'd expect it to show some loss.
IOW, I think your experiment, at least as you're presenting it here, is flawed.
tracknut wrote:
Interesting situation - I doubt you're discovering that TIFF is losing information though. The second image, which you describe as "edited in TIF" is actually a jpeg file (which of course can introduce some loss). Not only that, it's been compressed pretty highly. So I'd expect it to show some loss.
IOW, I think your experiment, at least as you're presenting it here, is flawed.
Dave
Thanks for your input, but they are all jpegs, even the first one.
Image #1: Converted from RAW to jpeg then cropped.
Image #2: Converted from RAW to TIFF, edited, then edits undone, then saved several times, then converted to jpeg and cropped.
Image #3: Converted from RAW to jpeg, edited, then edits undone, saved several times, then cropped.
I guess what I'm saying is that all the images you're showing us are compressed jpegs. By definition they will show loss. So we're not seeing what you're seeing, if your claim is the TIFF images are losing quality. You'd need to post two images:
1) RAW->TIFF (1st generation save)
2) RAW->TIFF (after several generations of saves)
Neither of which you converted to jpeg anywhere, to show us what you're seeing.
By the way, you say "edited, then edits undone" in those two images. If it's not clear, when the image is "in photoshop" it is in a lossless binary store, so if those edit/undo exercises to not include saving and re-loading the image from a file, I don't think they're adding anything to this experiment.
tomrock wrote:
It's math -- for every pixel in a tiff, there is R, G and B data stored. End of story.
If you're seeing some degradation, it's because of something you're doing to the file before saving it. Are you using any form of noise reduction?
There has to be some logical explanation. Tiff is lossless.
No noise reduction. I am puzzled as well. The images posted are jpegs as they were all converted to jpeg as the final step.
In this test I edited my images and then undid the edits. I was always under the assumption that I could edit a TIFF file as much as I wanted without any image degredation. If this is true then the second image posted above should be identical to the first. This seems to show otherwise.
Image #2 - edited in TIFF (contrast, brightness, then undid these changes), then converted to jpeg, 100% crop.
I still see image degredation in the shingles of the second image....Am I missing something here, because my TIFF files are definitely being damaged in editing.
HerbChong wrote:
what are using to edit with? in Photoshop, once you have undone all your changes, it won't save the file because there aren't any changes to save.
Herb....
I am using PSP X. The only "permanent" change I made to each image was straightening, which I did to each image. I thus saved the image after doing this. It seems like editing and unediting the TIFF file damaged it somehow.
Here are the two images side by side. Again, the first is unedited and the second is edited in TIFF with edits undone.....
There is legislation being drawn up at this very moment.......anyone caught working in and repeatedly saving in jpgs will be subject to repeated waterboarding.
rhyder wrote:
There is legislation being drawn up at this very moment.......anyone caught working in and repeatedly saving in jpgs will be subject to repeated waterboarding.
Definitely true.
But this was edited all in TIFF and only converted to jpeg at the very last step. Thus, there should be no loss in IQ.
Most times in print it won't matter, but I'm planning on doing some big landscape prints and this type of image degredation in TIFF is a bit disturbing. I'm really stumped by this one.
As you probably know by now, JPEG is a lossy format. Regardless of what format (TIFF, PSD, etc.) your image was previously saved in, as soon as you save it as a JPEG, you've degraded the image.
The real test would be to open your raw image and crop a portion (as you have been doing), and then save it twice, once as a TIFF, and then as a JPEG. Open both images and compare the detail. If you have saved the TIFF as either uncompressed or LZW, you should have exactly what you started with, no loss of detail.
BobCollette wrote:
As you probably know by now, JPEG is a lossy format. Regardless of what format (TIFF, PSD, etc.) your image was previously saved in, as soon as you save it as a JPEG, you've degraded the image.
.....
True, but then both images should show equal degradation. Why does the second image show more degradation?
The only difference between these images is that the second image was edited as a TIFF file and the first was an united TIFF. Then both were converted to jpeg. Thus, if there was no damage to the edited TIFF file then both images should be identical.
TIFF records the RGB value for each pixel. JPG encodes 8x8 pixel blocks in a form of Lab notation. If you are familar with Lab space you know that the "a" and "b" channels which carry the blue/yellow and green/magenta color information look very low in contrast. That also makes them highly compressible with minimal loss of color fidelity. Most of the detail is carried in the L (luminance channel). How compressible that channel is depends on the content of the image, which is why JPG file size will vary.
The quality variable in JPG controls the extent to which similar values - Lab components, not RGB - are averaged and treated as the same tone. The fewer tones, the more tonal degradation. At level 10 there is very little averaging and as a result very little degradation. Some random noise might be spotted by pixel peeping at 100% in flat tonal areas like the sky or the surface of a car, but when prints from a level 10 jpg and TIFF file are compared side-by-side the variables of the printing mask any differences. Thats why most pro labs suggest file submission as level 10 jpgs. Levels above 11 or 12 significantly increase file size without any tangible, visible difference in the print at normal viewing distance.
What is equally important with respect to perceptual IQ is how the file is sharpened before being saved for output. Sharpening done after resizing and should be varied according to the output size, characteristics of the output device, and the viewing distances. It is impossible to judge sharpening for a print on screen, you need to make test prints on the actual printer being used with varying degrees of USM and look at them from the distance the image will be viewed. If making a large print use a small section for the test but view it at the distance the large print will be seen from to accurately gauge the perceptual effect of IQ. The color temp of the lighting where the print will be displayed should also be taken into account. The best evaluations are done in the place where the final print will hang, or identical lighting conditions.
But I still don't see how it explains the loss in IQ that occured from editing the TIFF file. These edits were all undone prior to converting to jpeg. Thus, I don't see why both images shouldn't look identical.
as far as i am concerned, any application that after undoing all changes thinks something is changed is broken. on top of which TIFF has many variants including JPEG compression. normal TIFF is lossless compression and unless you are saving in one of the lossless formats, all bets are off.
Herb...
michael49 wrote:
I am using PSP X. The only "permanent" change I made to each image was straightening, which I did to each image. I thus saved the image after doing this. It seems like editing and unediting the TIFF file damaged it somehow.
michael49 wrote:
Chuck - that information is very helpful, thanks.
But I still don't see how it explains the loss in IQ that occured from editing the TIFF file. These edits were all undone prior to converting to jpeg. Thus, I don't see why both images shouldn't look identical.
I suspect what you are seeing is an unintended consequence of the jpg conversion process which is resulting in the same thing unsharp masking does: create additional contrast on at the tonal transitions which creates the illusion of more sharpness.
You might try applying moderate sharpening to the RAW file before the conversion to TIFF and JPG and see if that makes a difference.
