Moldy, Oldies (Where's The Craphics?)

[Omega-TI]

After looking at a lot of these old TI newsletters online, and seeing tons of old .PCX, RLE and TIA format graphics in those old newsletters, I started wondering how many of you "Elder 99'ers" have old graphics hiding out on old disks just waiting to once again see the light of day, or possibly even made into a disk image to two for sharing?

 

What was your favorite graphics program? Did you just use TI-Artist or some another program generally lost to history and forgotten? The old graphics stuff is an aspect of our the hobby that I've not really seen mentioned much, it might be fun to explore it and see what's been hiding for decades.

[Tursi]

TI Artist + was my favorite by far - of all the software I actually bought for my TI, that got the most use (second probably being the Missing Link). I was also fond of GIF Mania although not very many images came out as well as I'd have liked. (Still, I kept disks and disks of GIF files that were tolerable. )

 

I actually converted a few way back in the day, using a STOS BASIC program I wrote for a friend's Atari ST (because STOS could save as some native format that I've forgotten, which could then be converted to GIF, which he then sent back to me). Nothing overly special though, especially today!

 

Let's see how loaded a post can get Most of these will be my crap art, with the few converted GIFs I still have... I didn't cheat, these were all TI Artist format before I started (converted using Thumbs Plus and my Photoshop plugin for reading them).

 

Four frames from my Missing Link TMNT game, and and abstract box.

 

Floppy power supply schematic (I used one I built from this for years), and Waterville Gang art (some of which ended up in Waterville Rescue).

 

Rescue end frame, 'Flipette' drawing, Super Space Acer title, and two game titles I never finished (COTM was a door game on my BBS though!)

 

Dolphin Quest never worked, though I spent time on it. Manhunt never happened, Texin was my handle for a while, brother's artwork, and more WG.

 

I was starting to build my own shell around the RAMdisk, and jealous of Geneve I used the swan. Also Super Sled Acer (with bad spelling intact!)

 

Mass Murderer was a cheesy little maze game I did in BASIC and c99, and ported to the Apple 2, and then later I changed the title page to something less evil and hex-edited the binary to make the text nicer.

 

And some GIFs that I kept converted.

 

This one is NSFW, so I spoiled it, but I doubt that there's enough to see in the monochrome dithered image to be worrisome, so I posted it anyway. The things we accepted 30 years ago!

 

 

 

 

[Tursi]

Just for comparison's sake, since I still have two of those GIFs, here's what my image converter can do today

 

[moocowmoo]

I've come across a number of disks with various graphics in the massive collection of 11602 disk images at ftp://ftp.whtech.com/emulators/pc99/pc99%20dsk%20collection/.

 

 

[+Vorticon]

Just for comparison's sake, since I still have two of those GIFs, here's what my image converter can do today

 

p1.pngp2.png

 

Pretty amazing...

I seem to vaguely recall that you had created a loader for XB to allow the display of these bitmaps from within that environment so that XB programmers could add fancy splash screens to their programs. Am I correct in this?

[moocowmoo]

Pretty amazing...

I seem to vaguely recall that you had created a loader for XB to allow the display of these bitmaps from within that environment so that XB programmers could add fancy splash screens to their programs. Am I correct in this?

Yes, his converter can create an XB program from a picture. After it shows the picture the program needs to RUN another program as pretty much anything else will cause extended basic to crash.

[Tursi]

What Moocowmoo said. It's just one of the export modes.

 

Also, since VDP is completely corrupted, the first thing your next program needs to do is usually clear the screen and call charset.

 

I'd add, that if you want XB with Bitmap graphics, the Missing Link is a better way to go.

Edited July 21, 2014 by Tursi

[+Vorticon]

Ah OK.

And yes TML is a truly powerful environment.

[+Vorticon]

Tursi, I am currently working on my next demo for the upcoming TI Faire in Chicago in November (no, I'm not telling, yet... ) and I have a question for you in your capacity as the graphics conversion guru here:

Taking a standard RGB image file (3 bytes per pixel for Red, Green and Blue values), what algorithm have you used to convert that data into one of the 16 available colors on the TI for each pixel? Now I realize that the TI only allows a single color for each set of 8 horizontal pixels, and I will tackle that problem later, so let's just hypothetically assume that I can assign each individual pixel a separate color.

[Asmusr]

A simple method is to choose the color from the palette with the smallest distance from the true color. The color distance can be defined as:

DIST = SQRT((R2-R1)^2 + (G2-G1)^2 + (B2-B1)^2)

There are other more sophisticated algorithms, but this works quite well.

[RobertLM78]

A simple method is to choose the color from the palette with the smallest distance from the true color. The color distance can be defined as:

DIST = SQRT((R2-R1)^2 + (G2-G1)^2 + (B2-B1)^2)

There are other more sophisticated algorithms, but this works quite well.

As a math guy, I'm interested to see some of these other methods .

[+Vorticon]

A simple method is to choose the color from the palette with the smallest distance from the true color. The color distance can be defined as:

DIST = SQRT((R2-R1)^2 + (G2-G1)^2 + (B2-B1)^2)

There are other more sophisticated algorithms, but this works quite well.

 

I apologize if I sound a bit thick, but I'm not quite understanding this formula. How do I obtain R1 and R2 values for example? Do they represent successive pixel red values? Also, once I have the distance, how do I apply it to the 16 color TI spectrum i.e where is the origin I am comparing that distance to?

[sometimes99er]

You have the source pixel. The color is RGB. Each letter being a byte. We called it R1 G1 B1. Then all of the 15 colors of the TI is the target (each one called R2 G2 B2). Which of these 15 colors comes closet to the source. You don't have to take SQRT to get the exact right distance. Just store all 15 results and pick the one with the lowest value.

 

[+Vorticon]

You have the source pixel. The color is RGB. Each letter being a byte. We called it R1 G1 B1. Then all of the 15 colors of the TI is the target (each one called R2 G2 B2). Which of these 15 colors comes closet to the source. You don't have to take SQRT to get the exact right distance. Just store all 15 results and pick the one with the lowest value.

 

 

Ah OK I get it. Thanks!

Now I guess I have to figure out the RGB value of each of the 15 colors. Any idea how to do that?

[sometimes99er]

It's a bit different for a few different emulators. Let's take the one used in Magellan (because it's the easiest to retrieve right now).

 

 

 

 

And the one from Classic99.

 

 

 

Edited July 27, 2014 by sometimes99er

[Asmusr]

As a math guy, I'm interested to see some of these other methods .

 

See http://en.wikipedia.org/wiki/Color_difference

 

 

This is a palette from js99er, which again was ripped from MESS:

    this.palette = [
        [0, 0, 0],
        [0, 0, 0],
        [33, 200, 66],
        [94, 220, 120],
        [84, 85, 237],
        [125, 118, 252],
        [212, 82, 77],
        [66, 235, 245],
        [252, 85, 84],
        [255, 121, 120],
        [212, 193, 84],
        [230, 206, 128],
        [33, 176, 59],
        [201, 91, 186],
        [204, 204, 204],
        [255, 255, 255]
    ];

[+Vorticon]

Thanks for the help guys I can move forward from there.

[Tursi]

That's not strictly the algorithm I use anymore. But the distance algorithm is tried and true (it's just the Pythagorean theorem ).

 

For a long time I got "better" results with the 'perceptual' matching -- the idea being that because the eyes are more sensitive to green and less sensitive to blue, you could scale the distances to skew the results without hurting the result too badly. All I did there was scale each axis in the 3D color cube before calculating distance (originally I used the average scale values for monochrome, multiplied by 100, so 30 times red, 59 times green, and 11 times blue. I later hand-tweaked the values a bit. It does result in some color shift but better preserves visible detail (sometimes).

 

The current version goes back to a straight color-cube distance comparison, but converts all the colors to the YUV space first. This emphasizes brightness as one of the three major axes rather than each individual color. The results seem better again - preserving detail with less color shift.

 

I have a half dozen palettes in the converter (though I only use the Classic99 one right now). Determining "actual" color is tricky, although many people have claimed to get it "right". The problem is that right is not only a component of the video chip analog components (which themselves are only approximated on the data sheet), but the output circuitry, and the tuning of the monitor itself. "Close enough" is probably as correct as it really gets. Still, I was trying to pin down an artist and have them do a color match for me. It's difficult because artists are flighty creatures and leg-hold traps are banned (I got as far as two colors, once, before she decided she really needed her Mac to do it well. ).

[+Vorticon]

How about taking a YUV raw image file and converting it to RGB then calculating the distance?

From Wikipedia:

 

 

Will this improve the rendering of the image?

[Plastik]

MANHUNT.png

 

Roli just reviewed a game called manhunt for the vectrex. I'd love to see a clone of it on the Ti, was there an original manhunt? Maybe one that Rockstar used as a basis for their manhunt game for the PS2?

[Tursi]

How about taking a YUV raw image file and converting it to RGB then calculating the distance?

From Wikipedia:

 

 

Will this improve the rendering of the image?

I don't see how it would.... you are starting with RGB anyway. But with the way our eyes work, my opinion is that YUV is a superior color space for matching.

 

It's really all opinion, though.

[TI-Sissy]

How about taking a YUV raw image file and converting it to RGB then calculating the distance?

From Wikipedia:

 

 

Will this improve the rendering of the image?

 

huh?

[RobertLM78]

 

huh?

Click on the 'jibberish' to see the formulas .

[sometimes99er]

"Close enough" is probably as correct as it really gets. Still, I was trying to pin down an artist and have them do a color match for me. It's difficult because artists are flighty creatures and leg-hold traps are banned (I got as far as two colors, once, before she decided she really needed her Mac to do it well. ).

Actually, I'd rather have 10 amateurs than 1 expert guesstimating what's "close enough". Mostly we'll end up with "amateurs" using the thing anyway (emulators etc.), so I guess their (amateurs) perspective weighs in heavily ?

 

 

PS. Since I don't have the real iron anymore (+/- 1986) I suppose I shouldn't vote on what I think is best, on the other hand, I lean heavily on graphics in this TI-99/4A area, and I've gotten pretty entangled with the MESS palette. Does that count for anything ? I guess even the F18A solution may influence perceptions ?

 

Edited July 28, 2014 by sometimes99er

[Tursi]

Actually, I'd rather have 10 amateurs than 1 expert guesstimating what's "close enough". Mostly we'll end up with "amateurs" using the thing anyway (emulators etc.), so I guess their (amateurs) perspective weighs in heavily ?

I have deficient color vision, which is why I have to rely on algorithms and the like for color matching. I also know a lot of people who work with color as part of their daily life. So I want to ask one of them to look at a TI palette on a television and say "here's my estimate". At that point I can look at it and see what I think. I've got a half dozen "correct" palettes for the 9918 already, and nobody has had to stop work because of that.