FreeDoom recognition characteristics

[wesleyjohnson]

Thank you Gez and Quasar.

I have also been considering scanning the known lighting sprites at load time to identify their flame locations.
It would require significant code with the usual image recognition problems.

This has several advantages though:
1. Does not require FreeDoom to have any additional lumps, nor does it rely upon FreeDoom members to update a special lump.
2. Would work on other old wads.
3. Can be customized to the particular engine, and what it is displaying (like a corona).

A scan for bright pixels
a. pixels where at least one color > 230, and total intensity > 400.
b. where more than half the neighbors are bright (eliminate outliers)
Find the center of the spots and cope with up to 3 (or 4) of them per sprite.

Information such as flickering, would have to be from knowing the sprite identity (torch, lamp, candle, etc.).
Strange wads that redefine such things would create problems.
Perhaps a dehacked extension that can change it.

The only requirement upon FreeDoom and other wads, would be that their flame shapes not get too strange, not overlap, have contiguous bright pixels, and limited to 3 (or 4) flames per sprite. I would hope that those requirements would be acceptable.

Anyone try this before ??
Any new rational can you add to support one or the other idea ??
I think it would help if the engines had some common requirement instead of each imposing individual support requirements upon new wads such as FreeDoom. ??

Thank You.