I actually asked Sony Support about that. Their reply: "We can confirm that with Lossless compressed RAW there is a minimal quality loss. To have no impact on image quality I suggest using Uncompressed RAW files." Lossless isn't what it used to be.
Their first reply was "we have passed your question to a higher technical team", then they came back four days later with the above reply. I was enquiring about the A7R mark V, which introduced the much needed "lossless" option. I think I asked because I wondered why they kept the uncompressed option and because experts warned that Sony did that before with "lossless" formats.
"lossless" has always referred to compression, not sampling - but it seems camera manufacturers want to change that for marketing reasons.
Similarly (without starting an audiophile thread): Recording a vinyl record and compressing to a MP3 is "perceptually lossless" but will be different to compressing to a FLAC, never mind that the sampling output will always have random noise.
It’s because manufacturers keep inventing their own formats instead of using plain DNG, and you can only figure out the difference reverse engineering the file. Even for the same manufacturer and file type, they may change encoding details between models or firmware revisions.
There are situations where they may decrease the bit depth of the final image if there’s not enough dynamic range, there are situations where even though the file is “uncompressed” the camera already does noise reduction and essentially compresses detail in the image, and so on…
What is the problem there? It essential provides an unlimited license any Adobe patents needed to implement DNG, and revocation if you assert a patent against Adobe related to DNG.
It provides an unlimited license specifically for reading and writing spec-compliant files, which could conceivably open you up to issues if you have implementation errors or attempt to extend the spec.
How is DNG tooling these days? It's been ages since I checked, but it used have near zero support in (FLOSS) tools (including ffmpeg/imagemagick). You even had to use Adobe's Windows converter.
DNG is well supported by the typical raw libraries in my experience. My Ricoh records DNG natively, and I work with the results in rawtherapee and darktable. I am less sure about ffmpeg or imagemagick as I don't really automate raw processing.
I see however that DNG support is still read-only-ish, e.g. libraw still can't convert to DNG [1]. Most camera's still output raw in their own formats, which may not and probably should not be what you store them as, long term.
The worst part is its often in areas people improperly understand the benefits and downsides.
Biggest problem with good raw compression is you have a linear DNG, half processed essentially. Great, the file size is smaller, but now you miss data that processes like AI denoise can benefit from as the image is already debayered.
On the flip side, good compression like DNG 1.7 spec's jpeg-xl compression is borderline magic.
Lossless is actually lossless.
The lossy flavour is so good even at 105 megapixels in 16 bit (per color channel) I would challenge anyone to spot a noticeable difference compared to the original, a file possibly 20x it's size.
On a tangent, bits per channel is yet another part people split hairs over. 14 vs 16 has almost no difference, no the colours are not 'better' even in a full 16 bit workflow, the only real world perceivable difference is your darkest darks are more precise and under extreme editing conditions do look a little better if being raised extensibly in post. I digress 16 is bigger than 14 and yay marketing.
Looping back to compression, 14 bit raws without compression are padded to 16 bit lengths due to word sizes and file constraints. This bit throws off the less technically minded who make all sorts of assumptions about file sizes and being 'more lightweight to edit'.
I'm kind of surprised we haven't seen things like 16 bit luminence, 12 bit chromanance. (I guess to do that before debayering would require RGBW pixels or something like that)
The problem is camera marketing makes it really unclear what you're actually getting. Really lossless? Does it trigger more on-body processing? Does it throw away bits to be "visually lossless"?
If you shoot a few thousand photos and then find you can't fix exposure as well as you'd hoped and the whole batch is worse, it's a pretty big disappointment, so it's smart to be risk adverse and skeptical.
IMHO marketing is almost always (part of) the problem. They shouldn't drink or smoke that stuff …
Equating "lossless" with "visually lossless" or some other phrase is newspeak. We could call a JPEG of quality >= 95 (or 96 .. 99) visually lossless too then.
Losslessness is easy to define: compress something, then uncompress it again and both the original and the uncompressed file should compute to the same (cryptographic) checksum.
These forum discussions seem unconnected to reality IMO. Sometimes funny, more often strange, like audiophile discussions about oxygen free copper cables. Why?
The Fuji X-T4 and X-S20 here produce images of 6240x4160 pixels, but I almost never look at images in 1:1. My 4K monitors, mostly set to 2K, display 2560 x 1440 pixels. And even when switch them to their full 4K resolution I don't view my images in 1:1 obviously. And the tablet l'm typing this comment on offers "meager" 1800 × 2880 pixels. Most family members look at images on their "smart" devices nowadays, where 2K or 4K aren't present. So even decompressed images are fine for them.
I have my cameras configured to take both JPEG (fine) + RAW (lossless) of course. Fuji JPEGs (and Canon ... etc too) are fine for most casual viewers. And if I want to crop certain parts, or adjust certain details (esp exposure), I have my RAW images as a fallback.
Storage? My SD cards are 256GB and my disks are definitely not a problem either.
I almost never look at full images in 1:1, either, but I find very often that I have the wrong lens for an application and/or the wrong exposure.
You don't have to go very far before you've cropped a 20 MP source image into a 4K or 2K image, and if that part of the image that you've wanted to highlight is not well-lit, well, exposure is logarithmic and I want all of that RAW color depth that the camera can find if I'm going to turn black or white into perceptually accurate colors.
It's true that when my framing and exposure are great out of the box, I probably wouldn't notice or care if JPEG compression cut my file size by a factor of 4...but that's not always the case.
> My 4K monitors, mostly set to 2K, display 2560 x 1440 pixels. And even when switch them to their full 4K resolution I don't view my images in 1:1 obviously.
macOS, but my understanding of Retina is that even if your effective resolution is lower (e.g., my 6K ProDisplay has a resolution of 3008x1692), applications can designate regions to be "original resolution," so I'm getting the image's original resolution in the editing window or region.
The main reason my wife shoots lossless is touch-up editing in light room. You can rescue some pretty crappy photos when all the sensor data is present.
Honestly the best argument for uncompressed is actually nothing to do with file quality or loss - it's that Apple only supports uncompressed Fuji RAWs.
You cannot preview or process lossless compressed Fuji RAWs on iOS natively but the uncompressed files are equal to Apple's own RAWs in support. On the field, it is sadly worth every byte to be able to grab a file directly off the camera and tweak it or send it to an editor. :/
The problem is that the size difference between compressed and uncompressed is enormous. This adds up quickly if you shoot sports or wildlife with 20fps bursts.
> Honestly the best argument for uncompressed is actually nothing to do with file quality or loss - it's that Apple only supports uncompressed Fuji RAWs.
