Recreating CRTs - NFGworld!

Reply · Quote

2009-04-24, 13:38 #16

Member since Apr 2009 · 15 posts
Group memberships: Citizens, Members

Show profile ·

Link to this post

In reply to post ID 1733

I'm trying to wrap my brain around the convolution algorithms used to make gaussian blur, to see if there's a way to work out a halation blur. Thinking out loud here; bear with me...

Convolution is roughly this:
[Image: http://img15.imageshack.us/img15/7460/convolutionfilter.gif]

[Image: http://img15.imageshack.us/img15/7460/convolutionfilter.gif]

Apparently blur works by taking the center pixel's color value and summing it with the adjacent pixels' values, weighted based on distance? Something like that.

Where standard blur has the same values at all times, e.g. (25%)(50%)(75%)(CENTER)(75%)(50%)(25%) we would want the values to be variable depending on the value of the center pixel. Thus if center pixel is 255 bright:
(25%)(50%)(75%)(255)(75%)(50%)(25%)
But if it's 128 bright then perhaps:
(9%)(19%)(38%)(128)(38%)(19%)(9%)
There's probably an optimal drop-off curve for any given brightness; determining that would take experimentation.

And rather than merely sum the adjacent pixels, we'd need to enlighten them realistically. That would also take experimentation.

Reply · Quote

2009-04-24, 17:51 #17

Member since Apr 2009 · 15 posts
Group memberships: Citizens, Members

Show profile ·

Link to this post

Oh, interesting. I was inspecting some arcade photos up close and realized that the screen in the example had its RGB subpixels arranged vertically instead of horizontally. So I tried doing that in Photoshop, and got some curious results...

Benefit #1: The harsh transition between blue and red subpixels acts as its own scanline.
[Image: http://img9.imageshack.us/img9/9331/phossimgenvgng.jpg]

[Image: http://img9.imageshack.us/img9/9331/phossimgenvgng.jpg]

Benefit #2: Does not need any brightness/contrast step, as it stays pretty balanced in both bright and dim scenes.

[Image: http://img16.imageshack.us/img16/2264/phossimgenvff6.th.jpg]

[Image: http://img144.imageshack.us/img144/832/710395finalfantasyiii00.th.png]

Problems:
• It's still too goopy/blurry due to the gaussian blur being uniform across all values.
Still desiring a halation blur in the future.
• I'd still like a little bit of horizontal RGB distinction between pixels.
That can be done next time I'm in the mood to play.
• I've lost the glowy brightness that I had in the original Seiken Densetsu 3 image.
It's been five or six years, but I seem to recall a blur pass of 6 or so, and then using 'Auto Contrast'.
I'll tool around and see if there's a better way to handle that.

Reply · Quote

(Administrator) 2009-04-24, 17:57 #18

User title: Top Chaos Wrangler
Member since Sep 2007 · 1747 posts · Location: Brisbane, AU
Group memberships: Administrators, Members

Show profile ·

Link to this post

I've been playing a bit with it also and haven't really found atotally acceptable solution.

Is it possible that the arcade screenshots you've got with a horizontal phosphor arrangement are vertical games on a rotated monitor? I've seen a lot of monitors and never seen one with horizontal phosphors.

[Image: http://nfggames.com/system/arcade/special.php/b-d/bp-5/y-kof97/z-10/x-What?/.png]

Reply · Quote

2009-04-24, 18:15 #19

Member since Apr 2009 · 15 posts
Group memberships: Citizens, Members

Show profile ·

Link to this post

TATE mode? Nope.

[Image: http://img25.imageshack.us/img25/2496/rgbsf3.jpg]

[Image: http://img13.imageshack.us/img13/2620/sf3xboxvgascanned101.th.jpg]

[Image: http://img24.imageshack.us/img24/2507/rgbforgotten.jpg]

[Image: http://img25.imageshack.us/img25/495/forgotteni.th.jpg]

Isn't that just weird? It's a first for me, too.

Reply · Quote

(Administrator) 2009-04-24, 18:54 #20

User title: Top Chaos Wrangler
Member since Sep 2007 · 1747 posts · Location: Brisbane, AU
Group memberships: Administrators, Members

Show profile ·

Link to this post

That 3rd Strike pic is definitely a standard phosphor arrangement. The horizontal lines are the scanning sweeps of the electron beams, and you can see the phosphor elements running vertically through them.

The other one's from Tobias' site, and it's described thusly:

Here's a final comparison shot: Forgotten worlds (Xbox, filtered, badly upscaled, run through a scan converter, photographed, resized and scaled again) vs. the same game (Mame, original res, no scaling at all besides the one in photoshop for the shot below). I really wouldn't call the Xbox one to the left so much worse...

There's not nearly enough detail to say that it's using horizontal phosphors, but I guarantee it's not. =)

Reply · Quote

2009-04-24, 19:11 #21

Member since Apr 2009 · 15 posts
Group memberships: Citizens, Members

Show profile ·

Link to this post

Yeah, both shots are from Tobias' site. They both may have been taken off the same screen; maybe the electron guns are just a little bit off. Either way, it got me to try rotating the RGB stacks, which yielded an interesting result. Happy accident.

Reply · Quote

2009-04-25, 02:42 #22

Member since Oct 2007 · 3 posts
Group memberships: Citizens, Denizens, Members

Show profile ·

Link to this post

Don't mind me jumping in here, hehe.

It seems to me that the reason you're having trouble getting a blur to look right, is that a blur is not what you are trying to reproduce -- you're trying to get a glow. Simplify the problem down to a single pixle, think vector monitor, Asteroids shots, something like that. You have a single bright point surrounded by nothing.

With no processing, you get a single bright point surrounded by nothing. There actually should be a little glow from the monitor anyway, but I know the effect you mean on those old CRTs.

Now if you blur that, you are spreading the light into the dark areas, yes, but you are also spreading DARK into the LIGHT areas. You only want to spread the light, not the dark!

It seems to me that rather than a blur, you want an additive filter, where you add a percentage of the source pixel to the surrounding ones. This way you are never "spreading dark". For a horizontal-only example, something like this:

10% 25% 50% SRC 50% 25% 10%

You would repeat for every pixel, and you process each color gun separately. Lots of tweaking would be needed - I don't know how much fall-off you'd want, and you'd amost certainly need to start with a darker than normal image, but that should give a better glow effect.

Now, how you do that in Photoshop, I'm less sure.

Reply · Quote

(Administrator) 2009-04-25, 08:12 #23

User title: Top Chaos Wrangler
Member since Sep 2007 · 1747 posts · Location: Brisbane, AU
Group memberships: Administrators, Members

Show profile ·

Link to this post

Tursi's absolutely right. I think Psiga and I had realized that but kind of had trouble working out how to do it... And still, that's the key innit? I think I'll look into the idea of making my own filter today. Can't be that hard, can it? <fear>

oh, completely unrelated, someone posted this 2600 CRT emulation link on my other forum, more or less out of the blue. It's related to our interests.

Reply · Quote

2009-04-25, 13:20 #24

Member since Apr 2009 · 15 posts
Group memberships: Citizens, Members

Show profile ·

Link to this post

Heh, funny. Someone linked me to that out of the blue as well. I think I really like their color bleed method, whatever it is.

Reply · Quote

2009-04-27, 02:09 #25

Member since Sep 2008 · 5 posts · Location: Munich, Germany
Group memberships: Citizens, Members

Show profile ·

Link to this post

Very interesting topic! If I had more time on hand right now, I'd join the Photoshop experiments. The SF3 shot and the left side of the Forgotten Worlds pic were taken from a 14" Commodore 1084S monitor with a real shitty 2MP digicam, so there's not much detail to be seen.

I'll try to experiment with live overlays once I get my new video capture card. Also waiting for my neighbour to finally get a flat screen tv, so I can grab his 29" Sony RGB-CRT....

When using emulators with natively running screens (1080p on the TV or UXGA on the PC) I always find it rather hard to find a proper setting for scanline overlay etc. To achieve the slight blur of the CRT I find myself using lower than native res and have the LCD upscale the picture. Depending on the display, this adds a very nice - slight - blur to the picture which makes the picture much more CRT-like. Still far from the great HDR experiments posted here, but a good approach nevertheless.

Tobias

Reply · Quote

2009-05-03, 23:59 #26

Member since Nov 2008 · 2 posts
Group memberships: Members

Show profile ·

Link to this post

Interesting topic...

Anyone knows BlueMSX? It's a win32-only open source MSX emulator, having an interesting set of configurable filters called BlueTV.

Reply · Quote

(Administrator) 2009-05-05, 16:19 #27

User title: Top Chaos Wrangler
Member since Sep 2007 · 1747 posts · Location: Brisbane, AU
Group memberships: Administrators, Members

Show profile ·

Link to this post

By way of MooglyGuy I found this rather intense refutation of the work mentioned in post #23:

Quote by MooglyGuy:
This effect has pretty much no fundamental basis in reality, it was just a bunch of people throwing full-screen effects at the wall and seeing which ones they liked the look of. It is no more an “emulation” of CRT effects than the PNG-based filters that you can use with MAME. Besides that, the effect just plain looks too crappy.

He then links to these amazing Blargg libraries, saying

Quote by MooglyGuy:
Folks, THIS is the only legitimate “emulation” of a TV signal and its output. NRS and blargg based their work on the hard, documented math equations that involve generating and displaying an NTSC video signal on a television.

Reply · Quote

2009-05-07, 00:38 #28

Member since Nov 2008 · 2 posts
Group memberships: Members

Show profile ·

Link to this post

So not PAL fun?

Blargg does very nice and Open Source stuff. He deserter lots more attention and help than he has.

Reply · Quote

2009-05-17, 20:50 #29

Member since Apr 2009 · 15 posts
Group memberships: Citizens, Members

Show profile ·

Link to this post

I did mention the blargg NTSC filters in my first post, yeah. It's serving a different purpose, and I'd still want a robust CRT simulation which captures the look of electron beams scanning over phosphors.

Which is different from the Photoshop noodlings that I've been posting; those would just be blitter effects that, as MooglyGuy says about that other thing from Georgia Tech, have "pretty much no fundamental basis in reality."

I did one last thing in Photoshop, but it's just getting silly now, all merging three layers and doing four or so separate passes of gaussian blur, plus adjusting contrast. And for what? I think it looks better than bilinear filtering, but with that many steps of filtration it had darned well better.

[Image: http://img30.imageshack.us/img30/5690/ffvitownphosvi.jpg]

[Image: http://img30.imageshack.us/img30/5690/ffvitownphosvi.jpg]

I'm pretty much done fooling around for now. This is just reminding me why I abandoned the gaussian blur method in the past. What we'd need for this to work is a blur filter that varies depending on brightness.

Reply · Quote

(Guest) 2009-05-21, 23:15 #30

No profile available.

Link to this post

Subject: Glow Filter

http://www.linuxtopia.org/online_books/graphics_tools/gimp…

I don't know if any of you are familiar with the GIMP, but if you want to see what the effect of a filter such as you have described would be, the GIMP is free to download and install, and the plugin referenced above is included in the latest Windows installer.

It looks to me like it is doing pretty much exactly what you were suggesting - I created a test image, half black, half white, with a dot in the middle of the opposite color, and ran the filter against it. On the white half, the white invaded the black dot, but there was no change in the color of the area that was already white. On the black half, the white dot gained a halo, but again, there was no alteration of the color of the white dot itself.

In other words, in contrast to a gaussian blur, where the border becomes fuzzy, in the Soft Glow case, the border remained sharp and defined.

I did not find any Photoshop filters with an identical effect, though I only did a quick search; given the number of Photoshop filters in existence, an exhaustive search would be genuinely exhausting. I did find a short tutorial that looks like it might be having a similar result, however:

http://www.photoshopessentials.com/photo-effects/soft-focus/

Howard, the Grum