Image format and embedded color profile change on upload

We could program all those things into the resize function. But the code wouldn’t be beautiful anymore. As you have to use different options for different fileformats to tell imagemagick not to do any conversion other than “resize”.

And somehow he didn’t like my suggestion to only allow one fileformat. :stuck_out_tongue:

Are you being purposefully obtuse? How about outside the support of browsers? :slight_smile:

Or you can just use PNG24:outfile.png as I noted. Either way it’ll keep the colorspace as desired, which is good! :smiley:

Good policy to follow. How about HEIF support? :roll_eyes::rofl:

I think someone mentioned something about being able to directly upload a file to pixls instead of to someplace on the web - does this mean there is a way to upload a file from one’s own computer to somewhere on the pixls website?, somewhat like bugzilla allows to upload files by navigating to one’s local copy of the file? If so, what’s the procedure? Is there a help page somewhere?

  1. If it’s in the context of a threaded discussion here, the best option is to zip the file up and attach it to a post.

  2. If it’s a resource for wider consumption you can also add it to the github org. (Not sure which repo to use for something like this yet, but we can figure something out)

  3. If it’s not appropriate for git, we can also add it to the files section of the site (I’d have to dig up the credentials for posting it, but if you want it I’ll grab it for you).

That’s the question. After the file is zipped up, what is the next step? Upload it to the web and provide a link? Or is there a way to navigate to the file on one’s own computer and upload it directly from the computer?

1 Like

Oh for the love of pete. You ever get doing so many things at once you screw up some of them? That’s me right now. I’m sorry (in my head it sounds simple, but I’m not being clear). :frowning:

You can upload the file right here in the forum - just upload directly to your post using the upload button:

image

Or you can just drag and drop from your machine into your post composer.

I apologize for spacing out!

I was being silly too. I removed my other post because I thought @patdavid had answered your question. Just drag and drop into the post editor. Proof:

dragndrop.zip (124 Bytes)

1 Like

Possibly, sometimes I can’t tell :slight_smile: . But which browsers? And what does “support” mean? Clearly Firefox “supports” linear gamma RGB matrix profiles, but the images don’t display properly in the shadows. But just about all RGB matrix profiles with more or less perceptually uniform TRCs do display properly, assuming the right choices are made in about:config.

I’m fairly sure that Firefox supports some LUT profiles, but not others, perhaps LUT profiles with the XZY Profile Connection Space are not supported? I think that’s correct, but it’s been a long time since I last checked.

And what about V2 vs V4? Firefox only supports V4 profiles if the right switches are set in about:config.

Are AdobeRGB-so-called-compatible profiles made using the default LCMS D65 white point and not compensating for hexadecimal rounding “weird” or “not weird”, “custom” or “not custom”? They don’t match Adobe’s actual specifications, and they aren’t 100% well-behaved, which is one of the defining criteria for an RGB working space. But these are the types of "AdobeRGB-compatible profiles distributed by most free/libre profile makers.

OK, cutting to the chase, I’m guessing that by “not weird” and “not custom” you might mean “more or less closely approximating the actual profile specs for sRGB or AdobeRGB ICC profiles, including the specfied more or less perceptually uniform TRCs”.

But that does leave to one side the issue of V2 vs V4, and also the issue of actual browser support, so perhaps really the only profiles that aren’t weird or custom are V2 sRGB profiles?

  • Take the top 4-5 browsers in use, in their default configurations.
  • Check their color management capability.
    • Cross-correlate the results to arrive at a color profile supported by all.

There you have it. My initial guess would be sRGB v2, but you’re far more of a color expert than I am.

Adapt for the least-available that provides the ability to support the largest segment of your audience if at all possible. This is just being cognizant of the media used to share images (in this case a browser that you can’t know which one it is, on a web forum). You decide how you want to share your images. :slight_smile:

Well, of course V2 sRGB is the lowest common denominator, and that’s the color space I always recommend for anyone sharing images on the web unless they know the intended audience is equipped to properly view larger and/or V4 color spaces on properly calibrated and profiled monitors using properly color-managed browsers.

Which is to say, almost no-one except maybe people who participate on photography forums, and even then a lot of people won’t meet all the listed “unless” caveats. I don’t meet all the caveats because my monitor’s color gamut barely exceeds sRGB, though I can see a nice swath of greens and blue-greens that fall outside the sRGB color gamut.

But the specific request was made that I supply a test image that can be used to check whether image viewers can properly display linear gamma images. So I supplied two linear gamma 16-bit images, both pngs, both images had R=G=B everywhere. One image uploaded without any problems, the other was resized, converted to an 8-bit jpeg, and also converted to a RawTherapee version of sRGB, which left people asking the natural questions “Are you sure you uploaded a png? Are you sure the image isn’t already posterized?” And so on.

Supplying these two test images took a fair amount of time and thought to figure out what actually would constitute proper test images, and then to prepare them and upload them. I appreciate the fact that @patdavid has taken the time to investigate what went wrong and how to fix the problem, which apparently can’t be fixed. So that’s good to know. In the future I’ll provide zip files.

A bit off-topic: concerning file types. What is the purpose of accepting certain ones and not others? I am sure that whatever is on the white list is something that the back end knows how to handle and also should be able to remove unknown embedded tags or code that could potentially be malicious. However, the acceptance of archives like zip might still present a risk depending on what people place inside of them. Just thinking out loud. I know nothing about this stuff :slight_smile:.

PS I guess another reason from the top of my mind would be to prevent irrelevant file formats from clogging up the forums.

We allow a certain set of files that I originally chose to facilitate sharing in good faith. So image formats and raw formats as I am made aware of them. Also things like dtstyle and pp3 files, as well as plain text, patches, and zip/7zip archives. Last year I also decided to bump the maximum filesize allowed to 100MB per file (to accommodate bigger files). This also helps to avoid orphaned files if someone removes the file from it’s original location.

The purpose is to help avoid becoming a general dumping ground of any filetype at all. To at least try to keep it on-topic. Yes, there is always going to be a risk of accepting archives, but generally everyone has been good about not abusing this (and there’s almost always someone around if it does turn up something malicious).

It’s a balancing act - I want to make sure it’s as easy and frictionless as possible to communicate with each other and share relevant material. At the same time, we want to keep bad actors from abusing the system. We’ve been lucky so far with regards to that (and vigilant).

1 Like

Out of curiosity, is there a reason the image was not set to grayscale when it clearly is? Can you try posting the image again, but with the Color Type set to grayscale? (or a color version?). That way we can see if it was purely a problem due to the (incorrectly?) labeled color type?

My apologies, but I don’t know what you are talking about. I didn’t upload any grayscale images. I uploaded black and white RGB images. Some software in the upload process converted the original RGB image to grayscale, and then left the originally embedded RGB ICC profile still in place.

You might ask why didn’t I turn these images to grayscale before uploading as that way they’d be smaller images for uploading/downloading. But that’s a different question. The answer is that I don’t know how widespread support is for color-managed grayscale images in image viewers or even in web browsers. As the goal was test images for checking the display of linear gamma RGB images in image browsers, complicating the testing procedure by uploading grayscale images seems like a bad idea.

Again, neither image I uploaded was a grayscale image until after the upload process did whatever it did. An RGB image has three channels (ignoring the possibility of an alpha channel, which my images don’t have). A grayscale image has only one channel (again ignoring possible alpha channels).

An ICC profile for a grayscale image doesn’t have primaries. It only has a TRC. I uploaded black and white RGB images, three channels, with R=G=B everywhere, and with properly embedded RGB ICC profiles.

1 Like

Oh, that little icon with the upward-pointing arrow! Thanks! No need to apologize! It never occurred to me to experiment with clicking that icon (truthfully I never even noticed it) or with dragging and dropping files.

I’ll try the little upload icon, or else drag and drop, just as soon as I put together a zip file with the two test images, plus a README and an sRGB screenshot showing correct and incorrect display examples.

Does this allow outputting a 16-bit RGB image? A linear gamma 8-bit image will have posterized shadows.

Good point. For your case I think you’d want PNG48:outfile.png to maintain 16-bit. (just tested here).

Oh, very good point here - there is a significant distinction that needs to be maintained regarding single-channel grayscale vs “black and white” three-channel images.

I do a fair bit of black and white PP (probably because I don’t understand color… :smiley:), and I do so in three-channel R=G=B images. That way, if I decide to subsequently tone the image to, say, represent a particular photo paper, I have the channels with which to do that. E.g., my attempt at Sepia:

5-sepia

1 Like

That’ a very nice train photograph, imho. Toning and recoloring are the main reasons I keep black and white images as RGB.

Did you ever see a photograph on the web and then want to find it again but it seems to have disappeared? My “missing photograph” is close-up of an older train engine, maybe a quarter of the way around from straight on view, in an urban train depot-type place, black and white of course. Every now and again I search for it, still haven’t found it again. The photographer was someone well-regarded as a photographer, but I can’t remember his name.