HACKER Q&A
📣 jacobwilliamroy

Where do I go to talk to people about color matching?


So at work I wrote some python to match colors in photoshop. Basically it takes a CMYK color, an attribute (lightness, hue, chroma) a direction (-1 or +1) and a scale factor between 0 and 1 and it picks a certain number of colors from a database for which only that attribute is modified. The good news is that it works, but there are two problems: 1) it's really slow taking around 2-3 minutes and 2) it produces a lot of garbage which right now I work around by just increasing the number of colors for it to pick, which does not make it go slower. I figure it's time to ask for help. Where do I go to talk to people about this? Is there like an IRC channel or a forum for graphic design? I know Onyx Graphics has a product that does a similar thing and it's a lot smaller and faster than my solution but it doesn't work in every case.

  👤 coldtea Accepted Answer ✓
How does "it works" go together with "it produces a lot of garbage"?

I understand that it also produces some correct (or correct looking) results, but the code sounds like it has lots of false positives from what you're saying, so it's probably not the correct algorithm.

As for the time it takes, it doesn't sound like something that should take so much time, but not sure how it's in the code.

I'd look for these things:

Is the DB storing the color info as CMYK? If so, maybe go through your existing table and add another table or extra columns with pre-computed columns for lightness, hue and chroma.

Then you can get a CMYK value as input, translate it to lightness, hue and chroma, and select only the rows where the query attribute (e.g. lightness) is different to the one you got. You can even apply the scale factor in the SQL query directly (e.g.

  WHERE 
  CASE 
    WHEN $direction = 1 THEN lightness > $lightness + ($scale_ratio * $lightness)
    ELSE lightness < $lightness - ($scale_ratio * $lightness)
  END
Is the DB the slow part? Do you even need the DB? Perhaps you can loop through the colors (creating each in memory), and if you need to lookup some unrelated attribute like an assigned per-row id, you could query the DB for that after you found the N "matching" colors, and only for those.

Can you express the distance algorithm in SQL (or a PL/SQL function)? Is it faster?

👤 gus_massa
What happens if there is no match in the database?

My guess is that you are replacing each pixel in an image with the "best" color in the database. Is this correct? Can you post a good and a bad sample image?