Talk:Raw image format

Edit request

This silly article claims that negatives are "not directly usable as an image". Really? I guess all of the negatives available for sale, as art, and all of the movies using negatives as part of the movie exist in some other Universe? I'd suggest the editors of this piece actually think about what they are writing and not expand analogies into the ridiculous. ... Oh, I found another gross blunder: it seems that Android 5.0 - according to this article - allows the smartphone user to take RAW images...how does a camera "take" a file format? It should be obvious that any device "takes" sensor dependent data - the only question is how are those (voltage? (for CCDs)) measurements encoded (into a file) for storage? (with possibly several stages of processing (intermediate representations of the raw measurements) prior to storage in a datafile, of course.)216.96.79.240 (talk) 21:35, 21 October 2015 (UTC)

That statement about negatives referred to an exposed but undeveloped film, on which the image is invisible, rather than a developed and visible negative. The use of the verb "[to] take [a picture]" in the context of photography alludes by convention to the entire photographic process, rather than to any one step in that process as distinct from it. 47.208.130.197 (talk) 14:31, 3 May 2025 (UTC)

reversible steps

I'm not sure this is accurate some of the processes are not fully revisable as they involve rounding of the numbers which will prevent going back. You certainly cannot go from the final image back to the original data. The entire process induces a loss when converting to any of the more common formats.--Kitchen Knife (talk) 14:25, 22 February 2021 (UTC)

Yeah, definitely wrong, the raw-to-viewable process is completely destructive. I think whoever wrote that was trying to infer the vacuous truth that the original file isn't altered when you run the converter to output another format, but worded it unconditionally to the point of confusion. Fixed. 47.208.130.197 (talk) 14:31, 3 May 2025 (UTC)

"ARW 2.3.2" listed at Redirects for discussion

The redirect ARW 2.3.2 has been listed at redirects for discussion to determine whether its use and function meets the redirect guidelines. Readers of this page are welcome to comment on this redirect at Wikipedia:Redirects for discussion/Log/2024 August 26 § ARW 2.3.2 until a consensus is reached. 1234qwer1234qwer4 17:18, 26 August 2024 (UTC)

JPEG, TIFF, RGB, as linguistic standins

There were many ambiguous references throughout the article to JPEG as RGB, compressed, 8BPC, a file format, and the assumed target format of raw conversion; to RGB, or to 8BPC, or to SDR, as the assumed target format when converting from raw; and to images stored as compressed vs. uncompressed, lossy vs. perceptually lossless vs. truly lossless.

References to generic JPEG can be especially misleading. It is neither internally RGB (but as Darkfred noted above, actually YUV even when encoding RGB sources,) nor externally restricted to RGB inputs (CMYK, for instance, is also commonly used by Photoshop.) And while usually in the EXIF/JFIF file format, also occurs as a codec elsewhere, possibly including (if I'm properly understanding some of what I've read elsewhere) in raw files not just for preconverted previews, but in whole or part to compress raw sensor data. The one mention of the lossless "JPEG" format also opens the door to other obscure features depending on implementation and who you think the governing body of JPEG is today, such as 12-bit channels and HDR, even when excluding explicitly distinct JPEG standards (LS, XL, XR, 2K.)

RGB, aside from being far from the only color representation targeted by raw conversion, is itself neither a very precise description (separate reconversion from raw can be done even for other RGB colorspaces such as sRGB vs. rec.2020 for maximum precision,) nor an entirely stable one at the moment given the current transition to deep color and nonlinear/HDR (themselves presently in flux as standards) even for user software in which images are now consumed and edited. Demosaicing may cease to be a safe assumption too, given OLED subpixel geometries abandoning strict RGB stripes.

Breezy mentions of compression can also easily mislead here, in no small part because "raw file" has long been in many computing contexts an explicit synonym not of "unprocessed data" but specifically of "uncompressed data." But also because of many weaselly industry uses of "lossless" to mean "near lossless"/"perceptually lossless," and the often unstated use of various types of compression inside poorly documented photographic raw formats, image-centric, generic datastream, and both.

For all of these reasons, I settled on most often using "TIFF" in the article to refer to a generic conversion target intended for viewing, orthogonal to any destructive compression or color representation. This is, of course, also somewhat more ambiguous than I would prefer, especially since the uncompressed raster format inside TIFF I'm referring to (rather than the container file also shared with most raw formats) has no name, and all of TIFF's competitors (Pict, BMP, Targa, etc.) have died away. The best alternative I could think of, "viewable/viewing format," seemed too wordy.

This reduced the amount of strict technical falsehoods in the article, but is still IMHO inelegant and ambiguous. 47.208.130.197 (talk) 14:31, 3 May 2025 (UTC)