_Astara: click "Respond" to post comments

Confusion: how can a post be labeled with screening AND over-filtered? They are direct opposites, no?

It seems to me (???) that if some people feel it has screening effects in it (which should be few if any, but won't say I didn't
miss things), and **other** (???) people think it is over-filtered, doesn't that put it squarely between two opposites?

It seems to imply that different viewers on the same "board" [yande.re], use conflicting values -- how to resolve?

I wish the people who had one or the other of the views could talk and discuss what is really desired in posts so that posters wouldn't get conflicted feedback, but, ideally, feedback that would tend to 'herd posters' submissions towards "some, non-conflicting, desired" standard.

Probably not the place to have a discussion on this in comments, but thought I'd mention it since this scan seems to be a representative case.

Thanks for any feedback though, even if it is a bit confusing...
;-)

First of all, no one tagged this as "screening" besides yourself..

And this one is over filtered as hell. I believe the "screening"'d old version (post #303825) is hundreds better than this one so I undeleted it.

A post can be overfiltered in most areas while still have screening left in some areas.
On this particular scan the edges and lineart definitely seem overfiltered while there are many artifacts in the black areas.

WtfCakes said:
A post can be overfiltered in most areas while still have screening left in some areas.
On this particular scan the edges and lineart definitely seem overfiltered while there are many artifacts in the black areas.

Yeah I noticed that later. I thought this one was based on the old one but seems not.

Anyway, this was sharpened so hard that even the colors were changed.

fireattack said:
Anyway, this was sharpened so hard that even the colors were changed.

That's a legitimate deletion reason.

The strange part is this one has details which literally do not exist in post #303825 which cannot be explained by sharpening alone. This makes me suspect post #303825 was denoised to an even greater extent yet with an 'artistic' filter that gave it that texture, rather than that being actual screening. Honestly both are pretty bad, but the blurry oversharpening of this one make it appear subjectively worse.

Astara if these are your personal scans, I'd recommend uploading an unfiltered version (300-600 dpi png) as raw_scan and let someone else make an attempt at filtering it.

Cyberbeing said:
The strange part is this one has details which literally do not exist in post #303825 which cannot be explained by sharpening alone. This makes me suspect post #303825 was denoised to an even greater extent yet with an 'artistic' filter that gave it that texture, rather than that being actual screening. Honestly both are pretty bad, but the blurry oversharpening of this one make it appear subjectively worse.

Astara if these are your personal scans, I'd recommend uploading an unfiltered version (300-600 dpi png) as raw_scan and let someone else make an attempt at filtering it.

Nothing artistic I'd say, they're just two independent filtered scan based on perhaps one or two raw(s) (which is not uploaded). The most obvious evidence is there are some "bright points" on her right arm (and apparently belong to original image), which doesn't exist on post #303825 (I believe the editor considered them as noises at that time but changed his mind when editing this).

What I meant by that, was that I've seen some third-party Photoshop denoise filters which can cause loss of detail while leaving a noticeable texture very similar to post #303825 when filtering sources with heavy screening. One of the filters in the Topaz Labs set I believe does this kind of complex denoising and texture restoration which is ill-suited to images with screening.

Though it would also make sense as well if those details were manually removed by mistake. Didn't think of that.

This one was supposed to have been processed less.

The original has less sharpness if I remember.

I can post the png (if it will fit size wise). But I think I'll have to
rescan it. Only 3rd party noise filter in PS was Noise Ninja.

But certainly no problem if someone else wants to do a better job.

Scanner is a standard Epson XL10000, w/48 bit color.

It doesn't scan directly to png... but to tif, but I usually find the png's don't retain the correct color balance.

BTW speaking of color balance -- that was a difference of balance. The original has a yellow color cast that colors even the black areas. I reduced the cast (not removing it entirely, but it as it was, it was a bit too much that resulted in a bit overall fogging of the image).

If you see color has changed when converting from tiff to png, then your scanner must be attaching an ICC color profile to the image. If this is the case, you'd need to convert the image to sRGB before saving the png.

If you are sure your scanned tiff is sRGB yet you are still seeing a color change when converting to png, then it seems likely your image viewer or web browser is not applying color management to png images.

In Firefox you'd need to enable the about:config option gfx.color_management.mode=1 to force images and other webpage elements without ICC profiles to be color managed as sRGB.

I'd still recommend uploading an unfiltered raw_scan png, so we could get a better idea of what your source scan looks like.

If you are still having issues with color changes when converting from tiff to png, you may as well just upload the raw tiff itself to some file host and link it here in the comments for the time being.

Well, ** most viewers** have problems with custom color profiles, unless you choose to limit your colors to the sRGB profile, which looks noticeably washed out on my monitor, compared to say, Adobe RGB(98), or NTSC or HDTV. sRGB was created for devices in the late 90's. It was an 'ok' standard for 10-15 years ago, but by today's standards, it's a bit limiting.

Views included: Explorer (thumbs and its built-in image viewer), FF (note -- the setting for gfx color manage mode is set to '2' (use it if it is there), percept=0 and a profile created by a color-meas device (Spyder3). 'FastImageViewerPro'.

To see diffs between Adobe 98, and sRGB, see the
http://en.wikipedia.org/wiki/SRGB#mediaviewer/File:CIE1931xy_gamut_comparison.svg

My monitor is *roughly* equiv to the Adobe curve and shows better blue and red response.

I think I got the png posted... with an sRGB profile, but that's not what i'd use for a high quality image.

Quoting wikipedia: "an sRGB image is often regarded as satisfactory for home use. ... Images intended for professional printing via a fully color-managed workflow, e.g. prepress output, sometimes use another color space such as Adobe RGB (1998), which allows for a wider gamut. ... but they also mention that one can convert a higher quality profile to sRGB for targeting consumer level devices...

Posted the raw scan as a child of this image being discussed -- feel free to keep whatever version you want, I like the 2nd image, but if I can improve on it, i'm all 'ears'... ;-) (if someone else does, I hope they explain their process -- I'd like to put out stuff people like vs. stuff that gets a 'C' or doesn't pass at all...

By 'washed' if you're exclusively talking about scans, i don't think it's about the gamut of color profile though..

I doubt it's 'washed' out in the general term. If you are using a wide gamut monitor, then you're getting over saturation, which would cause the issue. Those wide gamut backlit monitors generally have an sRGB emulation function though.
It has nothing to do with the color profile.

Well if the image is converted to sRGB without enabling "Black Point Compensation", that would usually cause the image to be washed out.

@Astara
I believe yande.re strips ICC profiles from jpegs along with other metadata, so you'd always need to upload images as sRGB for them to be displayed properly. ICC profiles embedded in png appear to be kept, and I believe Firefox defaults may handle properly but no idea about other browsers.

There are 2 ways to embed profiles in png. 1 is supported by 'many' viewers -- and that's just a "bit" in the header that says to use the 'standard' sRGB. I'm not sure when, but I get the impression it went in later, was the ability to add a full custom ICC profile. That's the one that is more problematic as far as creating.

On konichan, I upped an image in PNG, that I jumped through hoops to get the right color profile on, but after it was uploaded, it looked overly red -- i.e. it looked like the custom profile was stripped or ignored and it was using sRGB for display in my FF browser. When I uploaded it as jpg, earlier, it looked right. But here's the thing -- I can strip my 'custom' profile off of jpgs or tiffs and they will still look right on my monitor, but that's because the viewing conditions and color balance in the image are the same.

With png's what may be happening is that the sRGB profile bit is getting set on most of the ones I see, but they were not converted to an sRGB profile. Thus they will look washed out on newer monitors (which are *tending* toward supporting more bits -- especially if they are HDMI or 4K compatible). Using an sRGB profile is *similar* (but not as destructive) as limiting your image to "16-240" for RG&B vs. using 0-255 -- i.e. sRGB reduces your dynamic color range (but not in a linear manner like a simple
lopping off the low and high ends).

@DarkRoseofHell -- yes, my TV-monitor has the option to disable 10-bit color as well as limit output to older SDTV standards. My main monitor has options to use the older SDTV standards as well, but, it would be a waste of bandwidth to post such images.

@Cyberbeing -- if yandre strips off profiles, maybe they should just convert them to GIF's while they are at it.
See the image fidelity difference in a simple image like this:
http://www.color.org/version4html.xalter

If the profile is stripped, the resulting image may look nothing like the original.

A few other pages showing the effect of different profiles (same images -- only profiles changed... you can see the profile can make the image look completely different).

http://www.colorwiki.com/wiki/Stunt_Profiles
http://petapixel.com/2012/06/25/is-your-browser-color-managed/
---
The 2nd URL shows that even among modern browsers, some
may show the color of the car completely inaccurately by ignoring or stripping the color profile.

Just out of curiosity, are you working in design or something?

I still wonder why you would think sRGB images are washed even with a wide gamut monitor. Aren't common images today still sRGB and designed for displaying in sRGB? How many 'wider gamut' materials would you meet during daily usage?

At the moment, wide gamut is only useful in printing for PC related stuff.

TVs have a different color gamut however, but the source material is also different.

I have an opposite problem where if I don't strip my scanner's color profile from my scans, the images will look too dark in web browsers for me.

@fireattack -- see http://www.cambridgeincolour.com/tutorials/sRGB-AdobeRGB1998.htm
---
<Is it possible to have images show up 'inline' in discussion here? >

The green spectrum in about the middle of the page shows the problem.

If I have an image that spans 0-255 in green going from black to full green, and then if I apply an sRGB profile to it, the sRGB profile drops off the extreme lows and highs and focuses it's bits on smaller gradations, say in the 16-240 (not sure of exact numbers -- an example) range. So now when I display the same image on my monitor, the darkest black now looks like a very dark grey and the brightest green is muted, not using the highest (most intense) portions of the green... same for the other colors. My monitor isn't exactly Adobe or NTSC, but has similar coverage -- just shifted a bit. But the sRGB profile doesn't use the extremes because they don't reproduce well *in print*. However, even on a monitor you can see the difference between #000 and #111 or #EEE vs. #FFF. Most people's *monitors*, can display, *at least*, dark black, and may (or may not) be able to display the more intense colors outside the sRGB range.

If you look at Dell's new monitors (2k price range a bit high right now, BUT) they have 4K-5K (4x-6x HDMI) pixels and 10-bit color. I *can* scan in @ 16-bits, but I usually downsample to 8 before I post, since most people (including me) don't have a monitor that displays >8bits (10-bit HDMI displays excepted). sRGB was chosen as most common for mass-media net images -- not high quality ones. Primarily, it was meant to handle the limits of printer tech (not monitor tech). Most medium level monitors (500-1000) have the ability to show a wider gamut than sRGB. Unless you are printing these images a wider gamut will allow a wider contrast range than sRGB.

I hope between that page, and my explanation that gives some idea of why sRGB images look washed out on my monitor -- because they are limited to a smaller color range.

@WTF -- Do you know what your scanner and monitor's 'gamma' are set to? If the gamma on the scanner is < your monitor's gamma, you'll get the effect you describe.

Maybe your scanning SW allows changing the scan gamma? It might be set for the older Mac standard of 1.8, vs. the PC standard using something closer to 2.2-2.4. Just guessing though as there are more variables in this area than I can really keep track of.

You didn't get my point, I totally understand that the range differences between different standards and sRGB's is very limited compared to others.

However, since most of images are still designed for sRGB, which means they would look the best/accurate (in term of artists' intention) when displayed in sRGB range. Stretch them into full range of, say Adobe RGB, might look "better" due to the nature of people like high contrast more, is no way better than in-sRGB-range one in my opinion.