JPEG XL

It’d be cool if smartphones could do this focus racking thing

Call it ‘Deep Focus’ or smth

smartphones have automatic Photoshop. did you hear about the drawn moon and Samsung

I mean... You can notch the focus manually with the volume slider on most

wtf

oh you meant phone

idk phone camera is just megaoixels

nothing more

cleartype?

Chrome’s default rendering on Windows, whichever that is

ClearType is disabled at the OS level (with MacType installed and activated instead), but AFAIK Chrome doesn’t care about that

lol

iirc it uses cleartype but a hardcoded non default setting

edge has a fix for it to use the system setting

A Leech?

no, just a black caterpillar

Nice!

<:BlobYay:806132268186861619>

an old-ish one (July 2019)

What display is this

j1 2016

Doesn't the text look blurry with this weird arrangement ?

Isn't your smartphone the same?

As far as I know - no one oled/amoled screen has square pixels

this another smartphone. j3 2018

Not about the square pixels. But the arrangement.

Mine looks like this

how did you shoot these?

I don't think it's noticeable, even on this 480p screen.

the heavy vignetting suggests perhaps through a microscope of some sort?

I see this annoying pattern even at 400ppi 🥲

I have a webcam that has a focus lens, I don't know how they were able to do that, but if you flip it over the tip of the lens makes a perfect macro so I just use a little webcam lens that I flipped over and put my smartphone on it

LCD?

yep

my old super-tft

modern people know it as IPS (first)

in fact, when people start arguing about which monitor is better, naming dozens of technologies that are literally the same thing, I always ask -- listen, what's the difference between super-tft and IPS?

oled

amoled

see the difference? ✍️

I think this photo is the most revealing

the alignment pattern of those blue subpixels is interesting

it's still a square grid of pixels with one each of RGB per pixel though, just the B is positioned a bit irregularly and has a different shape than the R and G

while this one looks more like a camera Bayer pattern, with twice as much G as R and B

I'd almost have said they've allocated more light to blue to compensate for us being less perceptive to it. In other words reverse Bayer

I think blue oleds are just less efficient (less brightness per area) so they make them bigger to compensate

So strangely I was almost right, just with the efficiency of the light rather than the eye

Yeah, correct. The PenTile arrangement was common with early Samsung phones, but as they scaled the size of OLEDs up they started doing this with blue pixels

I had the 4k AMOLED Galaxy Chromebook, it had a similar pattern. The Samsung Galaxy Books that came out with OLEDs were similar too

🤏 👌

iPhone SE (2020)

IPS, not OLED

How did you take this photo

with https://www.aliexpress.com/item/1005004836002281.html (not strictly this model but one close to it)

I have just measured that at maximum magnification, it seems that roughly 2mm fills the width of the frame

which seems consistent with the ppi figures I can find online for that phone

the quality isn’t amazing though

at full resolution, one can see interesting blocking artifacts

and the image kind of looks upsampled

(my advice if using a device like that: go slowly when adjusting the built-in lamp’s brightness, it can get really bright)

eh, downscaling 4× in each direction and reupscaling to the original size hardly makes a visible difference at all

“12MP”, suure

(I’m using the comparison tool in libjxl to check)

to get a better understanding of what’s going on, I should probably try this (1) on something that one would expect to have more detail, and (2) at lower magnifications as well

perhaps even try to see to what extent multi-shot “super-resolution” techniques improve the results

(by slightly moving the subject manually)

my PC monitor

Mine's the same as that, although my picture is blurrier so I won't bother posting it

do you have a microscope too?

No I just have a Pixel 7 & a low PPI display lol

yeah, no

and the aspect ratio seems off as well

https://sami.photo/pano/hohentwiel/

and now, one of my favourite places in Switzerland: https://sami.photo/pano/golzernsee/

fun with projections

coolest infographic I've seen in a while: https://petapixel.com/2013/07/01/how-they-sent-photos-across-the-ocean-back-in-1926/

https://petapixel.com/assets/uploads/2013/07/pIqE68N.jpg

an infographic from 1926

reminds me this gem : https://www.youtube.com/watch?v=cLUD_NGE370

This is a good test image for testing Moiré in downscaling algorithms: https://www.pexels.com/photo/woman-in-black-and-white-striped-short-sleeve-shirt-and-blue-denim-jeans-3686769/

Found this one upside-down on the road in the sun. 😬 Didn't know if it was alive or not, didn't look damaged but was all scrunched up into their shell. Put them right-side up in the shade of a tree facing a pond. Checked a couple hours later and they are gone.

eaten

Hm, think I haven't posted these old ish vacation pictures here yet ~~totally didn't lose all pictures and just found these in an old Discord DM~~

Nice quality in low light! What did you take these photos with?

with a Canon RF 70-200mm f/2.8

It depends. Fujifilm X-T100's kit lens (XC 15-45mm ƒ/3.5-5.6) does a good enough job in a decently sized package, so small I can put the entire kit (not without a bit of trouble, though) in my jacket.

What did you stack with?

Let me guess, RAW developed in Lightroom? Have you tried any of that deep learning denoising stuff, either DxO DeepPRIME or (more recent) Lightroom's built-in AI?

btw i avoid OLEDs because these burn off

IPS might be less vibrant but still good viewing angles, saturated colors and no burnoffs

Zerene Stacker, but Helicon Focus would probably also work here

...Lumia Refocus? 🙂

in DxO PhotoLab, but I tend to stick to the basic denoising methods

“HQ” or at most “Prime” (the non-“deep” version)

I find that the AI-based ones are a bit too prone to artifacts to my liking

In those pictures you probably used PRIME?

actually just HQ (I’ve just checked)

the shots were taken at f/2.8 and 1/100s

Depends on the shot: 1) what was originally on the photo, 2) how strong is the noise. It usually works well, but yeah, a bad combo of these two can yield unexpected results, especially with the latest version, DeepPRIME XD.

Oddly enough I haven't got much luck with HQ on my Fuji, even in good light. I always use at least PRIME.

I have that camera. It takes blurry photos on moving subjects on all modes except sports mode. Is that normal?

does sports mode perhaps use a shorter exposure time?

from today’s zoo outing

What are your usual settings? I haven't noticed such behavior because I almost only shoot landscapes (and other stuff that doesn't move much)

Another thing is that

I bought my unit used from a woman on OLX.pl

Visually looking, it's been through some things in previous owner's hands

And while making test shots on a sunny day, I noticed that it's constantly overexposing

Turned out that the culprit was the mechanical shutter...

You can read the whole thing, it's not that long: https://www.dpreview.com/forums/thread/4644717

But here's a TL;DR anyway: **`m_swan`** > I bought a used X-T30 about a year ago on B&H. I've had a great time with it generally but lately have had an issue with the mechanical shutter. It seems to be lagging, staying open longer than it should. I noticed it initially one day shooting with auto-SS and everything was extremely over-exposed. After some experimenting I found that - with everything in manual - I could expose images fine in electronic shutter, but the same settings in mechanical shutter were very blown out. Interestingly - in burst mode it seems that just the first image of a set is overexposed, while the rest are OK. > > Has anybody experienced this sort of issue and had any luck resolving ? The high end of the repair estimate is nearing what I paid originally, and where I live means I have to send it off for a while. So just looking for ideas before I do that. A part of me hopes there's some really dumb user error I'm not accounting for. At this point though I've reset the camera with no improvement. > > Appreciate any thoughts! **`ziemek99`** > OP, did you fix this issue? I have exactly the same problem with my used X-T100, to the letter. Yes, that includes the 6 FPS burst, where first photo is overexposed and all the following ones are exposed correctly – I could reproduce that. On slower 3 FPS burst all photos are overexposed, though. **`ArUdd`** > Hi - just wondering if you managed to resolve your issue. I have the same problem on my x-pro2 now with completely different exposures between the MS and ES all settings being the same. I have been using the MS predominantly but noticed the overexposed images on my recent trip to Asia which ruined quite a few pic opportunities.

Mine is new. I don't remember the exact settings but it mostly takes blurry photos.

I use this lens. I don't know much about it

For landscape photography, it's fine though

Goats looking at the "Staff only" sign be like:

https://tenor.com/view/cant-read-gif-22252869

believe it or not but that image is stitched (just a bit)

hence the different filename pattern

I was going to send the two original images but got a notice that they were rejected because they were detected as “explicit” 🙄

well, here they are: https://imgur.com/a/2TnfkMv

the second photo makes up most of the stitched image but I missed the top of the left goat’s tail

what the hell, imgur thinks so too

people can have complex relationships with goats, I guess? xD

Well, my last phototour 🙂

holy

moly

https://cdn.discordapp.com/attachments/1073799013465333830/1106653160442961961/PXL_20220821_153013815.jpg https://cdn.discordapp.com/attachments/1073799013465333830/1106653160929509407/PXL_20220821_162135537.jpg I wish I took these with a better camera <:FeelsSadMan:808221433243107338>

still very nice tho

what matters is that you captured the moment

simply having a camera is more than anyone should ask for

what did you take them with? it’s quite nice already

my guess is a google pixel?

oh, hadn't seen the filenames 😁

nope <:CatSmile:805382488293244929> <@604964375924834314> Xiaomi Mi Note 10 I changed the stock OS to LineAgeOS, which was signed with Pixel keys so Google camera thinks I have a one <:KekDog:805390049033191445>

I'll never buy Xiaomi again, got so many problems with the stock OS

Red admiral butterfly

just got a new keyboard _and_ a new lens; trying out the latter on the former

more testing

Spider-mario spider pics when? <:thinkies:854271204411572236>

maybe when I find a diffuser I’m happy with 😁

https://www.amazon.com/Angler-PSFD-100-Portable-Speedlight-Diffuser/dp/B005ZSJCTC/ seems nice but doesn’t ship to Switzerland

maybe I’ll have it shipped to my parents’ and grab it when I visit them, or use one of those package reexpedition services

Aren't there are similar products available though?

you would think so, but I haven’t found one yet

sometimes, I’m almost tempted to use an A4 sheet of paper with a hole in it as a substitute

That's surprising tbh. Would expect something simple like that to be available everywhere

so would I

next test the former on the latter <:Thonk:805904896879493180>

Cuyahoga Valley National Park

<:thinkies:895863009820414004>

Is it raining outside?

it's old water treatment complex

Very old

Bot still work

nice rays of sunshine <:BlobYay:806132268186861619>

<:Hmmm:654081052108652544> Looks more brighter and have more contrast on my phone could it be because of HDR ? <:Hmmm:654081052108652544>

I checked the image file and it doesn't look to be a HDR image

so it's probably just a difference of screens

it seems to have an extended XMP but I can’t seem to be able to extract it

ah, got it

just one big `GCamera:hdrp_makernote` blob in base64

Shot on the Pixel 7 RAW, edited in Darktable

playing in blender (not sure if that counts as photography… I did shoot the background, it’s https://sami.photo/pano/golzernsee/)

Hmmm, wonder how JXL fares with HDRIs...

for some reason, djxl fails to write back an EXR file… let’s see what it could be

``` lib\extras\enc\exr.cc:96: JXL_FAILURE: Unsupported pixel format for OpenEXR output ```

https://tenor.com/view/wtf-confused-question-marks-gif-12181142

wait, I thought I’d fixed a similar bug some time ago

fixed in https://github.com/libjxl/libjxl/pull/2565

Wonder what this is

It's an insect.

https://en.m.wikipedia.org/wiki/Mimas_tiliae

I think they are in the middle of reproducing in your nice photograph

Yeah, they are <:KekDog:805390049033191445>

But thanks, this seems to be the one

Moth porn on the Jpeg XL discord? More likely than you think

<:UwU:911376902411919421>

new toy

(I’ve also spent some time on https://www.microscope.healthcare.nikon.com/microtools/resolution-calculator/ to estimate the quality of the pictures I’ll be able to take once I receive my camera adapter)

(we’ll see if reality matches my estimations)

for now, I have a basic USB camera (5.7×4.28mm or ‘1/2.5" ’ sensor with 5MP) that came with the microscope, but I haven’t tried it yet

Oh no... I got more pillows <:KekDog:884736660376535040>

Hm, seems like my phone's camera has some dead pixels which are very noticeable in the bottom right <:RaysCrying:686219918122942500>

At least I'll get a better phone in like half a year anyways with hopefully a 1" sensor instead of 1/2.55" <:KekDog:884736660376535040>

How is foot treating you? I wanted to test it out sometime

it's okay. p minimal, can do bg transparency, can do fonts i want

Hello. Do you know any app that uses this android 13 feature so i can get real HDR I can't find anything acceptable, google is wasted with copypasta news and playstore has no real hdr apps Yes, my phone has android 13 Poco x5 pro 5g to be specific No such a thing in miui camera https://developer.android.com/training/camera2/hdr-video-capture

opening HDR from the usual Files works on my sony (for raws)

youtube also works

“Everyone welcome Except cycles”

I am about capturing HDR like iPhone dolby vision

YouTube hdr works well and with bigger great nits

16 cores?

yeh

5800X3D

so 8phys 16virt

I used Fujifilm x-t100 to take that photo.

about bokeh, this video is just amazing, if you didn't know it : https://www.youtube.com/watch?v=9cT0jXI7l4E

is there a way to convert Apple HEIF images into Jpeg XL while keeping the HDR Tonemap and metadata?

I'd assume one way would be extracting the metadata to a file and then reapplying to the JXL during encoding. Not good for more than 1 image at a time though

but how would i get HDR to work?

im not too sure how HEIC encodes imaeges, although that might make more sense in <#805176455658733570>

it's just metadata

It can be set manually in cjxl

`--dec-hints=color_space=RGB_D65_DCI_Rel_PeQ` for example, but it completely depends on the source metadata

Does anyone know how to transcode Ultra HDR images from the Pixel to JPEG-XL, preserving the HDR quality & metadata?

does lossless jpeg bitstream reconstruction not in fact preserve the original bitstream in those cases?

I think they mean have the JXL file display as HDR, since it would interpret the HDR data as just normal metadata instead

I do mean that. I'd like to take Ultra HDR JPEG -> HDR JXL

Manually setting it to a HDR colorspace might work as mentioned above, but it would take some encoder work to do natively

it doesn't preserve the tonemap, but technically that's a second jpeg

I can open an issue, because this seems like an interesting feature

And a very widespread one being integrated into Android

ah, working with JXL’s native HDR support would involve applying the gain map

the gain map is what takes an “ultra HDR” image from SDR to HDR

simply tagging the regular image, which is SDR, as HDR would not give the desired result

I feared that

maybe a flag like `--gainmap-retention=1` might be useful

or `--gainmap-apply=1`

is it not embedded as metadata? does `cjxl` drop it altogether?

I can send you a sample

had to put in a .zst to prevent discord shenanigans

here's the JXL, where it isn't preserved

Moving to the thread so this channel doesn't get crowded

is there a tool that can convert an "ultra HDR" jpeg to a 16-bit png in some HDR space (so where the gain map has been applied)?

perhaps it would make sense to make cjxl recognize ultra HDR jpegs and make it transcode them to actual HDR jxl

I had been wondering whether to try my hands at some Ultra HDR manipulation tools as a sort of holiday project, also to gain some deeper understanding of how it works, but haven’t actually done that so far

it shouldn't be too hard

AFAIU it's pretty much a multiply in linear space

> In the case that the Gain Map is scaled to a different resolution from the primary image for storage, the sampling method must be bilinear or better, and is implementation defined.

“or better” 😁

better according to what criterion?

what

implementation defined resampling? ugh

why do people do that

what resolution is used in practice for the gain map? is it typically 1:1 or is it typically subsampled?

often 1:4 iirc

I assume it wouldn't be as simple as layering the two jpegs, upsampling to match resolution, then set to multiply and flag the JXL as HDR

it wouldn't be that as it should be in linear space

but otherwise...

doing 4x upsampling in an undefined way ("bilinear or better") and then use that as _multipliers_ for pixel values sounds like a horrible idea

I consider gain maps something that system engineers can come up when planning on a white board

Terrible kludge to avoid fixing things at a fundamental algorithmic level

I don't think it will ever work in a way that is acceptable and will be forgotten as a bad chapter in image compression in 5 years

I anticipate quality, interoperability, inefficiency (density, coding speed) and security problems

agree on everything but the security problems, not sure how you'd get more than you already have

well it does make it possible to make an image that renders completely differently in SDR than in HDR — this could in principle be abused for e.g. spear phishing that targets specifically people with an HDR screen. I don't think it's a huge issue but still...

I’d be all for that. Preserving the HDR is what is important when working with these images, *not* preserving the gainmap, imo

Embedding "If you can read this you spent too much money" in all HDR images from now on <:YEP:808828808127971399>

example of the security problem: you have an agreement that you review online on an HDR screen, where you agree to buy a house for 11000. When you print it the system uses the SDR version, you sign it without carefully rereading (since you read it on the screen), and the printed version has a price of 111000.

I think usually people mean something else with security problem, and it's not like it would be hard to do that even without (@media print css selectors), but I do see the point

yes, it is in the category of 'information system security' rather than 'computer security'

where 'information system' includes both computers and humans operating them

"From a sociotechnical perspective, information systems are composed by four components: task, people, structure (or roles), and technology." ((I learned such terminology originally in Finnish around 1985 from an 'automated data processing' dictionary that I won as a prize in a national coding competition... looks like the use has been relatively consistent and the same in English...))

that's why I sign only bitonal images, like faxes

When you have access to the html/css/js, it is easy enough to do all kinds of things. If you can do something with just an image file though, it means that anything accepting user-uploaded images is potentially affected, which is a much larger attack surface than websites accepting arbitrary user-uploaded html/css/js. It's for that reason that often the set of allowed upload formats is limited (e.g. only jpeg/png/gif since those are 'safe', but no svg or pdf because who knows how they will render) and/or uploaded images get transcoded from pixels to 'sanitize' them.

https://www.bram.us/2021/12/17/heres-a-png-that-will-show-a-different-image-in-apple-software/ this kind of thing, Apple fixed this as soon as they learned about it

https://isp.page/news/leica-enters-the-mobile-market-the-new-leica-cell-phone/

what

wonder how similar to that image it will turn out to be

don't they already have a phone? https://www.youtube.com/watch?v=DGWM5uYSlMA

I think that website is an AI content farm. The author seems to be made up, all photos of her online are clearly AI generated and don't even look similar.

ah, could be

We need a httpsh protocol where pages are written by certified humans 🙂

That reminded me, supposedly my phone has a Leica camera, or at least that's what's printed under the glass. Certainly doesn't hold up after 7 years and AI being crammed in though :P

how would that work?

Lol I dunno how to do it without creating privacy issues and centralized authorities. And it still wouldn't prevent certified humans to use AI tools...

yeah it would have to be something like c2pa++

https://www.instagram.com/reel/C3e4c5RtimA/?igsh=MTdkdGQ4dXJ5dG00bA==

Instagram believing my ASD would be lowering By 1,percent with tjat

😆😆😆

I am depenalizing this content,

https://www.facebook.com/share/v/ceDrRkZriCgJ4i7U/

Emma marro,

Probably I will instantly have a ban from ByteDance don't care 😆😆

Also Is important that the algorithm needs to be tuned for northic things

USA people can't see what's happening in Napoli

anyone got a galaxy s24? seems theres some issue with their jxl implementation of dng 1.7 in promode/expertraw

It's only while watching this that I remembered my phone has a dedicated 24MP Monochrome sensor https://youtu.be/4wf3ebnqzn0

Maybe I should try storing it's images as greyscale rather than RGB sometime

that's definitely an interesting choice for a camera sensor

It was an old Leica collaboration, dual sensors, one 20MP colour and 24MP monochrome. Annoying the monochrome sensor is only accessable though the default camera app, and can only take jpeg

Supposedly people have modified the firmware or guessed hardware IDs to make it take RAW, but over a weekend I couldn't get it to work

https://www.facebook.com/share/UBUSjN74xWFBQqEr/

Sorgenia released official Environment and climate benchmarks

08march 2024

reminds me of the glorious days of the yota phone

I do wonder what its sensitivity is like

Do you know if they use a UV/IR cut filter on the mono sensor?

Huh, interesting... Apparently it's a 12MP color sensor and 20MP monochrome, so it must be combining both for it's '20MP Color" mode

how can I properly convert (HDR) DNG to PPM ? `dcraw` is supposed to do it correctly, but default values give me yellowish output. Here a screenshot: left is DNG embedded preview, right is `dcraw -w` (which is closer to the actual look of the DNG but still darker) I want to automate for hundreds of file

The left looks like a jpeg preview embedded in the DNG to me

it is and I'm trying to search how to get the same tonemapping

I believe RawTherapee has a feature where it tries to set the parameters in a way that minimises the difference with the embedded preview

not sure how well that works

but could be worth a try

Oh nice didn't know about that software, thanks "auto-matched tone curve" definitely helped, needed a few adjustment (only thing I couldn't fix were the chroma noise) here's the result (43 MB lossless PNG - 31 MB lossless <:JXL:805850130203934781> 😎 - 5.8 MB -d1 <:JXL:805850130203934781> )

I thought there was a bot here allowing us to run `/embed image.jxl` 🤔 cannot find it

so here's the JPG

https://embed.moe/https://cdn.discordapp.com/attachments/803950138795622455/1218902957962170398/PXL_20240315_171208655lesschroma_lesswhite_d1.jxl?ex=66095ad7&is=65f6e5d7&hm=a5ed520c2644ec0b348f3e534789ec1c17a918a88d61b37a4ebe31fca512afb7&

Can I assume the settings I used to "remaster" this RAW will be the same for all RAWs of the same camera ?

For the birthday of Nicky jam I Imagined what if Niko Pandetta never existed in history? How would be the world of Image encoding? That's the result

Dua lipa photo

I also imagined a text before and after with Gemini 1.0 pro

Probably not without Libjxl 0.10.1 It would be pretty bad

I also did a study to comfirm that

In offtopic

Well this seems like a bit of a mess https://github.com/Beep6581/RawTherapee/issues/6273

Kurt cobain photo restoration

Chiara Ferragni photo restoration

https://photographylife.com/why-leveling-the-horizon-isnt-easy nice article

there are fun visual phenomena that cause the real horizon to be invisible, due to e.g. the sea reflecting the sky perfectly beyond a certain angle. Those are fun, though usually levelling according to the false horizon you see below the true one will be good enough.

_Pensive Pigeon_

(sounds like a release name for Ubuntu)

“Ubuntu 25.04 Pensive Pigeon”

A 7 year old phone and shaky hands aren't the best for long exposures, but managed to get... Something... Of the aurora last night, I think...

https://cdn.discordapp.com/attachments/498213582148730903/1238785391469924434/IMG_20240511_000835.jpg?ex=66408c4c&is=663f3acc&hm=3b275b4eea91e161cea8d2fea76d080ee515797a5cbb1ee56f160221265b6cd3&

nice

On the topic of 7 year old phone, here's the 2024 Eclipse captured by a 7 year old phone with a photolithographic plate

I can't find aurorae at 25N today

https://sami.photo/2024-05-11-aurores.html

Are there darktable enthusiasts here? I am not really familiar with neither darktable nor lightroom, but I noticed that darktable tends to have defaults that look rather dull, while lightroom's defaults look more vivid / contrasty. Does someone know if there's a darktable style or `darktablerc` file somewhere that makes it approximate lightroom's defaults a bit better?

last time I used it, there were two main types of workflows: the traditional one based on the “Base curve” module, and the newer one based on whatever the current iteration of the “filmic” module is

not sure which one is the default but the latter would tend to be dull until you tweak the filmic parameters appropriately

last time I used darktable it didn't support Sony ARW 4.0 still doesn't

Not sure if this helps, but I know Darktable does have its own internal colorspace it uses for certain modules that might explain some of the rationale behind its choices: https://eng.aurelienpierre.com/2022/02/color-saturation-control-for-the-21th-century/

> chroma is relative to the grey colour having the same brightness, no, same illumination, not same brightness

Color Appearance Models, 3rd edition: > Saturation is the colorfulness of a stimulus relative to its own brightness, while chroma is colorfulness relative to the brightness of a similarly illuminated area that appears white.

the chroma of a dark red patch on a white sheet of paper is relative to the brightness of the white of the paper; its saturation is relative to its own, lower brightness

https://www.dpreview.com/news/5796420320/affinity-announces-6-month-free-trial-for-affinity-photo

A JXL-friendly suite

Yeah

God is calling

There's something about the composition I like.

I like it a lot, just wish that one mast wasn't there

Same haha

anyone can check windows' Photos app with the Raw extension updated today?

some dng 1.7 photos using jxl for image (not the preview) seem to decode fine now, wondering if its a recent change

Ahh right, I remember seeing it update and was going to test but forgot

Thankfully, I was also going to test tinyDNG again too, so good timing...

I get nothing at all

You sure they're actually JXL DNGs?

The extension is based on libraw, and they only release updates once or twice a year, with the last one being in January

libraw updated its dng/jxl support after the current stable

but an issue with jxl previews/thumbnails of dng 1.7 encoded using jxl still exists so a snapshot around may/june

lemme check the encoding again

https://www.libraw.org/news/libraw-202403-snapshot

close enough, i guess the extension just rebased on that snapshot. as mentioned, dng 1.7 encodued with jxl and using jxl thumbnails/previews fail

I tried lossy JXL, lossy bayer JXL and lossless JXL with jpeg previews, then a lossless JXL with JXL preview just for good measure

None loaded

Is there even any software besides Adobe's which is currently capable of working with jxl dngs?

Libraw can be built with Adobe's DNG SDK, but no software includes that currently I think

Even ImageMagick has no support yet

Yeah, the only mention I could find was someone needing help from the libraw devs because the SDK was decoding it incorrectly, and ImageMagik's .NET page saying they don't know if they can include the SDK as a dependancy

Yes, Adobe's license for the dng sdk is a bit unclear

Which really does not help the situation

Oh, wait, apparently Adobe changed the DNG converter yet again and now it's making JXL previews, but how do I remove them or revert to jpeg...

Could you upload a file for me to test? Apparently it requires dark magic to get a file *without* being JXL everywhere now xD

Or actually, I guess you could see if this works. There's no JXL preview, but there is a jpeg thumbnail left over

this one doesnt open in photos with raw ext 2.2.2052.0

https://github.com/user-attachments/files/15585073/sample.lossy.dngs.zip

one should work other not

both reportedly load in adobe bridge

Yeah, the JXL file spins the loading icon for a while then gives up

I only just thought to check, but what Windows version are you on?

win10 current stable (19045.4717)

Hmm, I'm on 4651, I wonder if that's the difference

doubt, what fails on your end fails here too. whats odd is the dngs that work compared to before

id say that means raw extension rebased on libraw snapshot rather than the yearly stable

Sorry for keep going with this... So JXL DNG files now load for you? At least with standard jpeg previews... I tried making my own, but had to delete the preview entirely, which failed. And then downloaded the samples you sent, with the standard working as you'd expect and the JXL trying to load but failing

any tool best to recheck all dng metadata in detail?

dngs arent mine, looted off submitters specifically capturing using 1.7/jxl

I've just been using `exiftool -G1 -a Test.dng` which gets the job done for seeing the full resolution and reduced compression types

updated for jxl?

Hmm, the reason I brought this up again is because I just bumped into someone who shoots in JXL on their phone, maybe I'll ask them for a sample too

Yeah, most recent is probably best

odd there isnt some media kit for people to check this. jxl ran on generate your own files and thats troublesome for new formats that break existing conventions

the "generate your own file" is a recipe for chicken-and-egg why generate XYZ files if nothing supports XYZ why support XYZ if no XYZ files exist

quick question. is there any blocker to printers being able to print jxl images? like whats necessary for the ecosystem or needs an update

people never really think about this so i wondered if say a recent printer could print jxl images its sent or scan images to jxl

there’s no implicit reason, the format just needs more traction

If/When JXL becomes mainstream im presuming scanners will eventually scan in JXL, if not JPEG->JXL

where does adoption come from for printers? are there specs, does os need format decodable by some specific library somewhere? do modern printers decode on-device (ie for printers connected to and receiving images from the cloud)

the adoption comes from the engineer's manager who randomly decides to put it on the board and allocate working hours to implement it

jpeg xl support is probably 2 story points

maybe 3 if you want to tune the encoding options

^ Some holiday pictures taken when we were hiking in the Alps. Taken with my not-so-great smartphone (Motorola Edge 40, which I guess is a mid-range phone with no remarkable camera). The phone does allow you to shoot in DNG, so I had some fun with that in Lightroom to make some new HDR images.

nice river ! (discord fails to open them in jxl, despite the firefox plug in). Which part of the alps ?

around St.Moritz

I like that river image since you can see the glacier that feeds it in the background

Gotta get <:Cloudinary:806630933242445904> to sponsor you a high end phone for science™️ <:KekDog:884736660376535040>

The sunset yesterday as I saw it from my garden. Finally it is possible to create pictures that convey some of those deep orange and red colors in the clouds. At least on my laptop, with 4 stops of hdr headroom, it looks pretty nice — there's a ton of photon noise on the foreground as can be expected in these lighting conditions, focus is less than ideal, and I'm obviously not a photographer and my phone is obviously not a professional camera, but compared to what you can do in SDR with an image like this, it's still spectacular imo.

Totally agreed, looks amazing on my MacBook

New park nearby

<:This:805404376658739230> so 12068x10952 no works in discord lol

at the same time, the limit on opening in browsers chrome and firefox = 16k x 16k While the limits of the standard are 64k x 64k

Discord’s preview generator probably doesn’t support it, as Discord generates thumbnails

my AMD Athlon(tm) II X4 640 Processor 3.01 GHz can generate prewiew but discord cloud super-computer can not

It’s probably more it doesn’t want to

The upload preview is the full image resampled, the actual image preview in the client is a WebP, so probably an 8K limit

This is awesome!

Is it supposed to be pitch black in the dark areas though, or are you supposed to be able to see some details in there

I can see details in the darkest areas, although I don't have HDR

iPhone HDR; i mainly see photon noise but there is definitely some detail

our cat, Dame Blanche

close-up of a little flower

HDR looks really nice

This is a good question, cause if I open in Pixelmator Pro, I see lots of foreground. If I open in Affinity Photo, it's almost black.

Anyone have contacts at Affinity? I'm pretty sure they're handling HDR incorrectly for every file type

I can't believe the dynamic range off of a phone!

Yeah... it looks suspiciously dark. Like it's possibly getting displayed wrong... or maybe the wrong gamma info was written to the file.

One or the other... it looks really pretty but I think I’m probably supposed to be able to see more depth in the dark areas.

It's probably meant to be dark... but when displayed on screens that can't get really dark, it would be better if it pushes up the shadows just a little itsy bit, so they aren't totally blacked out.

Is that considered "tone mapping?"

Smoothing out the extremes so there isn't as much clipping?

Is that even something that color management systems are expected to do automatically?

bush

my youngest daughter in HDR

HDR looks really nice on iPhone

How do you edit these

I'm using Adobe Lightroom, since Cloudinary gives me a full Adobe suite license anyway so I might as well use it

I don't think darktable already has HDR editing in a way that allows you to actually see what you're doing

Does affinity photo work for HDR?

I bought a licence for graphics design

it's a world of difference between SDR and HDR, to me this feels like how I felt when going from 16-color EGA to 256-color VGA or when going from that to TrueColor. It brings a very noticeable improvement in fidelity.

I don't know but looking at <@594623456415449088> 's example above, it seems like it doesn't properly support it

getting HDR right involves more than just being able to work with higher bit depth

Being on SDR the main thing I notice is the noise, and thinking about just how well the JXL generated noise matches to these real examples

Or maybe a better description is the difference between a photo printed on regular office paper with a cheap inkjet printer, compared to a properly printed photo.

The JXL noise generation is meant to simulate photon noise, which is a physical thing that is just the nature of light. AVIF noise generation is more aimed at reproducing film grain, which is more of an artifact of analog technology from the past and only really useful for artistic retro style effects imo.

Can cameras denoise images then JXL photon nose add in ack

Theoretically yes, but denoising is a risky thing and hard to get right. There is signal hidden in the noise so aggressive denoising will remove a lot of information.

Is there any plausible use case for the photon noise for encoding from pixels?

Or is it more for when you add artificial noise?

If the actual ISO setting of the camera is known, it would in principle be possible to do a conservative denoising proportional to that, and add it back with noise generation. If only pixels are given, you first need to estimate the noise.

And I guess this has the advantages of if you don’t want the noise you can just tell the code to ignore the command to add the noise and then you’ve got a denoised image

If I recall the main idea was to hide compression artefacts, or increase density by adding it on encode instead of as a filter in the source program. It's also been used as dithering on 16bit files, trying to find the blog post of that...

So, like comfort noise?

For something like CGI, it certainly makes sense to add noise with codec-level noise generation instead of adding it to the image before passing it to the encoder

Because DCT really doesn’t like noise?

Actually, scratch that. Any comp compression doesn’t really like noise.

Never again... https://discord.com/channels/794206087879852103/803645746661425173/1266267110464688209

did jpeg xl end up helping a bit?

as the rough shape could be put in an MA tree and the noise could be stored raw but not need as many bits?

I can't find any trace of the blog apart from me mentioning it here months ago. I know it did have an effect and at practically no cost

The images attached show the results, e9 was 10 MB smaller than oxipng but no ISO noise was used or denoising

not even for iso noise, just for the better MA prediction

yes

at least on macOS

Not if You use Topaz Photo AI/Topaz Gigapixel AI/Topaz Denois AI. It can remove camera noise only.

most of the noise in most digital photos is photon shot noise rather than noise added by the camera

They have all the tools right there! I just think they're treating the data wrong when they open it. Cause I can easily adjust the sliders to get back to what it should look like… It's very strange since they brag about their HDR support… I gave up using it when I was trying to do the HDR images for the website.

Have been using Pixelmator Pro instead. It's HDR output options are awesome (besides the lack of JXL support). I've been going EXR -> JXL