Talk:JPEG

The article only has one sentence mention of gamma, and links to Gamma correction, but it seems that it could say a little more here. Specifically, what gamma value is commonly used in generating JPEG files. There is a mention of 8 bits vs. 11 bits, suggesting 11/8 or about 1.4, but it could give a little more detail. Gah4 (talk) 09:15, 21 March 2025 (UTC)

The article says that sRGB is commonly used. It has a two piece transfer curve that could be approximated with a gamma of 2.3. More details are in the sRGB article. You can find color profiles on Elle Stone's website to visualize the difference between the transfer functions on a dark low key image. -- J7n (talk) 01:20, 12 June 2025 (UTC)

Nobody would consider Q50 to be "high quality". It is barely usable. High quality in Photoshop is approximately 90. The lowest quality images on the web are around 60. The table would suggest that there is barely any difference for 50 levels.

A key benefit of Jpegli is support for higher bit depth when both the encoder and decoder are used together, and also more fine-grained quality levels in the useful range with the fractional distance parameter. -- J7n (talk) 01:32, 12 June 2025 (UTC)

“Arithmetic coding typically makes files about 5–7% smaller.[58]” is not reproducible, as almost nothing is known about the original files on which the measurements were performed; in particular, “typical” (a vague word) is not justified well in the cited paper. Suggestion: “Arithmetic coding makes certain files about 5–7% smaller [58], whereas certain other real-world files can be reduced by a factor exceeding 3.3 [links to http://bugs.debian.org/1127449 and http://bugs.debian.org/1007710], and certain test files by factors 13–655 [links to http://github.com/etemesi254/zune-image/issues/350 and http://bugs.debian.org/1067549].” — Preceding unsigned comment added by ~2026-13532-95 (talk) 18:22, 1 March 2026 (UTC)

Not familiar with image encoding at all, and this page does not help me with that. For the lengthy "JPEG Codec Example", there are a number of statements that would seem to constitute original research. In particular there are two categories of concern:

1. "Many of the options in the JPEG standard are not commonly used" -- this makes a statement about a standard format without a clear link to studies with usage frequencies of said features, or a statement by an expert in the field to back this. This kind of "inside baseball" is constant throughout the section, most notably with "However, some JPEG implementations in "highest quality" mode do not apply this step and instead keep the color information in the RGB color model," which sports a citation needed from February 2008.
2. "Due to the densities of color- and brightness-sensitive receptors in the human eye" -- while I appreciate the correct application of hyphenation, this is a scientific and potentially even medically-significant statement. I think this is particularly where the approach of describing an implementation seems to lose an encyclopedic tone. We can approach this in two ways:
   1. Link to an in-depth discussion or scientific research that was used to justify this claim. This would be less destructive to the current framing of the article, and implementors of the JPEG format (which this section is clearly intended for) would be able to make decisions based upon science.
   2. However, in general this format seems wildly out of place for wikipedia, even though I'm aware it's endemic in discussions of math or science. An encyclopedic framing would rephrase this in terms of the historical ways implementors encoded these color densities, and how science on this has evolved over time. If initial implementations were patented, did the patents include this research? Has this been a product differentiator?

The second point is especially significant because it makes a quantitative statement about the human eye, which is extremely surprising to see without a subsequent citation or simply a wikilink to the article on the eye. I don't know that research, but this is the kind of statement that can be misinterpreted by e.g. journalists or even scientists. At minimum, it seems to describe original research, and if it can't be cited for some reason, that itself seems to be extremely significant for a page which discusses patents and intellectual property at such length.

A final note is that the single citation in the "Discrete Cosine Transform" section (which disqualifies it from an "unreferenced section" tag) goes to a forum thread that does not remotely seem like an encyclopedic source: https://web.archive.org/web/20171017042422/http://forum.doom9.org/showthread.php?p=184647#post184647. I find this particularly of note because that precise link has the following response:

As you have IEEE material at your fingertips, perhaps you could check the source

N. Ahmed, T. Natarajan, and K. R. Rao, "Discrete Cosine Transform," IEEE Transactions on Computers, Vol. C-23, No. 1, January 1974, pp. 90-93

Why on earth is a wikipedia article citing a forum thread which itself tells the user to look up a real source? As above, if these sources are not available without a paywall, that itself seems like a very significant subject of discussion for wikipedia and particularly for this article which has such a specific analysis of patents on the format.

If the purpose of this section is to assemble the knowledge of the JPEG format that is otherwise paywalled, I'm all for that too--but then we really need to give implementors better tools than what this section provides.

It's deeply unclear to me how this conforms to wikipedia standards particularly around original research. I have added citation/clarify tags to several points mentioned above, but a discussion of how wikipedia generally incorporates technical articles with how-tos into the encyclopedic model would be greatly appreciated. Lcdrovers (talk) 22:56, 16 March 2026 (UTC)