Print

[Windy]

Windy

Tile Identification

As some people know, i was once working on a tileset optimizer, and on of the main problems with it, was the usability of the tileset, to combat that i created a method that would "compare" the side pixels of all of the tiles and see which ones would fit best together, the one issue with doing it that way was that while some tiles matched better than others, even though they weren't the tiles that were supposed to match up.

anyway.. i need some better ideas on how i should attempt it.

[Dayjo]

Dayjo

Re: Tile Identification

Is this for a tileset that is already split up, rather than like a "screenshot" of an area? If so, I suppose you could use the method you used before, but also do something like take an average of the colours that exist in the tiles and the ones with the closer averages should match better. However, that won't always work I suppose, due to the small size of tiles anyway.

www.dayjo.org

• 

[Windy]

Windy

Re: Tile Identification

yeah, already split up.  that was the suggestion i recieved first time round, however it wouldn't help increase the accuracy of the tileset

come to think of it, the problem tiles i have are with symetrical patterns, take the large lttp tree for example, if you split it in half, each side is a mirror image, and because of that, each mirrored piece matches up with its counterpart

a method i've been trying to develop, is one where it filters out distinctive lines which i figure i could use somehow.