The full version of Ken Fleisher’s recommendation, recently added to the text of this posting above, does acknowledge the benefit of TIFF over JPEG for greater-than-8-bit-depth images:

Higher bit-depth is really the only advantage of using TIFF over JPEG 10 or 12 (in terms of image quality).

The recommendation also says that JPEG is appropriate for terminal copies (my phrase, the text says “and if they do not require much pixel editing before use”). I think, in other words, no further transformations are expected.

For me, the key part of your reply was:

I feel that JPEG should not be disregarded as a delivery format because it�s perceived as “lossy”.

I, for one, and just coming to terms with the concept of “visually lossless” and its associated consequences as a practice for terminal (e.g. archival) copies. It is, of course, much easier to say “JPEG is a lossy compression method, therefore it should not be used for archival copies.” It is much harder to give the nuanced argument, as you did with your example comparing a TIFF with its corresponding JPEG quality 12 instance. (I almost used the word “derivative” instead of “instance” — demonstrating that I’m still struggling internally with the nuanced argument.)

I do not want to give a misleading impression to anyone who comes upon this post, so I’ve updated the text of the main posting with the full paragraph from the UPDIG site. Thanks for offering your perspective and setting the record straight.

“Archiving JPEG files: Conventional wisdom holds that the TIFF format holds a quality advantage over the JPEG format. This holds true only if the JPEG file is saved at less than 10 quality using the Photoshop standard. When using JPEG quality 10 or 12, the artifacts are either non-existent or insignificant. Higher bit-depth is really the only advantage of using TIFF over JPEG 10 or 12 (in terms of image quality). Some have argued that that JPEG, because of the way it encodes data, compromises color. This is a misconception. When using the highest quality settings, there is no loss of color fidelity. Therefore, if JPEG files are saved at 10-12 quality, and if they do not require much pixel editing before use, archiving JPEG files is not a bad concept, and it can save a lot of space. For many picture archives, the economics of storing large numbers of files dominates all other considerations, and JPEG offers a feasible solution to the problem.”

This relates the ideas that 1) the image is already edited or does not require significant editing, 2) that only the highest JPEG quality levels are used, and 3) that the solution is not for everybody but when appropriate, it can have certain advantages–namely significant space savings (translates to “cost savings” for most of us). Given these qualifying statements, I don’t feel the text requires any changes.

Also after re-reading the original message, I believe that for most professional uses, JPEG (with stated constraints) should be a viable solution. For scientific uses where introduction of any compounding error can distort experimental results, JPEG is clearly not a solution. But since these errors are so small (most are less than plus or minus 2), they simply will not be visible in most cases (large gamut color spaces noted as an exception, but even then the errors are only “slightly” visible and only if you look for them). I’m not saying that errors aren’t introduced, but they truly are insignificant.

That said, I do also feel that in some situations, though certainly not all, JPEG “could” be appropriate as an archival format as well. Take this example:

I took an image capture of a painting (color corrected, sharpened, etc., and then downsampled to 8-bit/channel). I saved one version as TIFF and one as JPEG quality 12, baseline optimized. I get the following results when I compare the difference between the two files:

17,020,752 total pixels in the image

(RED channel histogram values listed, but similar numbers for GREEN and BLUE channels)
8-BIT DIFFERENCE LEVEL – # Pixels
0 – 7,438,696
1 – 8,128,655
2 – 1,375,457
3 – 75,734
4 – 2,173
5 – 37
6 – 0
7 – 0
etc.

Typical camera noise for digital cameras is on the order of 2.5% and follows a gaussian distribution. (This is a gross over-simplification of noise, but is sufficient for this discussion.) That means we can expect 8-bit pixel values to vary by a maximum of plus or minus 6 pixels, with a standard deviation of about 1.5 – 2 pixels. In other words, based on noise characteristics, we can expect our image captures to have errors of plus or minus “2″ for 2/3 of the pixels, with errors no larger than 6. Although the example does not exactly fit this model, it should be clear that the errors introduced by saving as JPEG are very close to being on the order of typical image capture noise. Errors such as this will be visually lossless in most cases. An exception might be if using a colorspace such as ProPhoto RGB, then there may be “very” slight visual differences. Therefore, I believe that in some cases where it is not necessary to save the raw camera capture, JPEG can be used instead of TIFF for archival purposes. Stated another way, a JPEG file saved as above does not add any more distortion than typical camera noise. As long as the effect (which is admittedly compounded with the camera noise) is not visible, then why is this any worse of an archive file? I’ll stress again, this depends highly on the application and purpose of the archive you are creating, but I do see it as a plausible format for some cases.

- another key framing factor is the likelihood that in the coming decades, a growing number of born-digital images initially (and long-) managed by photographers will subsequently enter into public collections, and that encouragement towards better practices (when truly best practices may be unattainable) during those earlier phases of personal image stewardship potentially has wider, long-term benefits.

- one promising avenue for building connections between the library community and UPDIG may be the Preserving Digital Images initiative of the American Society of Media Photographers, undertaken with Library of Congress support via the National Digital Information Infrastructure and Preservation Program (NDIIPP); http://www.asmp.org/pdi/index.php has more on the “ASMP-PDI,” as it’s known.

- repository-focused work linked to UPDIG in some other key ways is happening under the auspices of ImageMuse (http://imagemuse.org).

And with that, I’ll stop for real. Promise.

Thanks for the comment, rob, and for pointing the authors of the UPDIG to this discussion.