MIF package v2.1

Page 4/4
1 | 2 | 3 |

By NYYRIKKI

Enlighted (5923)

NYYRIKKI's picture

03-01-2014, 12:33

Louthrax wrote:

For screen 2 & 4, I've no consistent algorithm to do that for now. Screen 10 palette correction was a bit complex, but not that much. Generating an optimal set of sprites (using OR color combination), with the 8 sprites per line constraints is a magnitude more tricky oO But I'm thinking about that, it's a very interesting math / computation problem Smile By the way, maybe some investigations have already been done in the past ?

I think it is more than just interesting math problem. It is very hard for program to decide what are the worst places that should be corrected... Even if it would be mathematically correct answer it might be wrong from user point of view... Usually places that are close to "golden selection" or that have more radical color changes should be first on the list, but that is not always the case... The fact that number of sprites is limited and there can be only 4/8 sprites on row makes this really hard.

I don't think there is much discussion about this, but check out this post & comments:
http://www.msx.org/news/development/en/screen-2-tutorial

Maybe this kind of problems could be solved with some "hint-painting" that would add the sprites in editor realtime for preview.

By Louthrax

Prophet (2414)

Louthrax's picture

03-01-2014, 12:51

NYYRIKKI wrote:

I think it is more than just interesting math problem. It is very hard for program to decide what are the worst places that should be corrected... Even if it would be mathematically correct answer it might be wrong from user point of view... Usually places that are close to "golden selection" or that have more radical color changes should be first on the list, but that is not always the case...

Yes, the mathematical approach is usually just... the simplest one Wink Achieving this first step for sprites-fix is already a challenge. For example, the CIEDE2000 algorithm, pointed by sd_snatcher, computes color-differences relative to the human-eye and gives very good results, but is very complex !

NYYRIKKI wrote:

The fact that number of sprites is limited and there can be only 4/8 sprites on row makes this really hard.
I don't think there is much discussion about this, but check out this post & comments:
http://www.msx.org/news/development/en/screen-2-tutorial
Maybe this kind of problems could be solved with some "hint-painting" that would add the sprites in editor realtime for preview.

Good idea, something like painting a mask over the image to indicate areas to patch in priority ? It could be a binary mask, or even a 0-255 priority mask, that would give a weight to the pixels to fix.

Page 4/4
1 | 2 | 3 |