Data is Beautiful

1185 readers

1 users here now

Be respectful

founded 5 months ago

MODERATORS

KnowledgeIsPower@mander.xyz

Beaver@lemmy.ca

WanderingPhoenix@mander.xyz

Analysis of 100 Years of Movie Posters (Around the World) (files.catbox.moe)

submitted 1 month ago by Blaze@sopuli.xyz to c/dataisbeautiful@mander.xyz

9 comments fedilink hide all child comments

Credits: https://old.reddit.com/r/dataisbeautiful/comments/1g7z2b2/oc_i_analyzed_100_years_of_movie_posters_around/

you are viewing a single comment's thread
view the rest of the comments

[–] Ephera@lemmy.ml 4 points 1 month ago (1 children)

Not a colorololologist either, but I'm guessing, you scan it and then calculate the average of the individual pixels. You can convert RGB color values to HSV color values to get the hue, saturation and value.

But calculating the average of the hue degrees (like was done for this post) is non-sense.
0° is red, as is 360°, so you get completely different averages, depending on how you calculate them:

(0°+120°+240°) / 3 → 120° (green)
(360°+120°+240°) / 3 → 240° (blue)

[–] nimpnin@sopuli.xyz 2 points 1 month ago (1 children)

Does hue make sense at all? If yes, you could calculate the vector mean (convert degrees to x and y axis, average, and convert back to degrees) and get something sensible out of that.

[–] Ephera@lemmy.ml 1 points 1 month ago

Ah yeah, good idea.

But good question, if it makes sense at all. Since hue is not a linear scale, the average is not exactly the 'middle' of anything. At best you could interpret it as the poster being somewhat biased in that direction, but ultimately you could have a poster with no green at all, which still averages out to green, and that just doesn't feel right.

A more complex graph, which shows the frequency of the colours used, might be more useful...