Created

Embed URL

HTTPS clone URL

SSH clone URL

You can clone with HTTPS or SSH.

Download Gist

Image Manipulation, Part 2

View kmeans_palette.R
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58
doInstall <- TRUE # Change to FALSE if you don't want packages installed.
toInstall <- c("ReadImages", "reshape", "ggplot2")
if(doInstall){install.packages(toInstall, repos = "http://cran.r-project.org")}
lapply(toInstall, library, character.only = TRUE)
 
# Image URL:
allImageURLs <- c("http://media.charlesleifer.com/blog/photos/thumbnails/akira_940x700.jpg",
"http://upload.wikimedia.org/wikipedia/commons/thumb/e/ec/Mona_Lisa%2C_by_Leonardo_da_Vinci%2C_from_C2RMF_retouched.jpg/402px-Mona_Lisa%2C_by_Leonardo_da_Vinci%2C_from_C2RMF_retouched.jpg",
"http://upload.wikimedia.org/wikipedia/commons/thumb/e/e9/Official_portrait_of_Barack_Obama.jpg/441px-Official_portrait_of_Barack_Obama.jpg",
"http://cache.boston.com/universal/site_graphics/blogs/bigpicture/obama_11_05/obama22_16604051.jpg",
"http://upload.wikimedia.org/wikipedia/commons/thumb/e/ea/Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg/758px-Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg",
"http://www.10mfh.com/wp-content/uploads/2011/09/dino_riders.jpg",
"http://images3.alphacoders.com/855/8557.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/ngm_101912/bp19.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/ngm_101912/bp26.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/ngm_101912/bp35.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/balloon/bp6.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/smithsonian_030512/bp14.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/smithsonian_030512/bp15.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/earth_day_2012/bp6.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/2011part2/bp1.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/2011part2/bp4.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/2011part2/bp15.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/2011part2/bp27.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/natural_world_2011/bp40.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/ngmphotocontest_111811/bp10.jpg",
"http://inapcache.boston.com/universal/site_graphics/blogs/bigpicture/ngmphotocontest_111811/bp54.jpg")
 
imageLoader <- function(url){ # This function takes a URL, and generates a data.frame with pixel locations and colors
# Download to disk, load
download.file(url, "tempPicture.jpg", mode = "wb") # Stash image locally
readImage <- read.jpeg("tempPicture.jpg")
longImage <- melt(readImage)
rgbImage <- reshape(longImage, timevar = "X3",
idvar = c("X1", "X2"), direction = "wide")
rgbImage$X1 <- -rgbImage$X1
return(rgbImage)
}
 
##########
# Part 2 # Identifying "dominant" colors with k-means
##########
 
rgbImage <- imageLoader(allImageURLs[5]) # Pick one, or use your own URL.
with(rgbImage, plot(X2, X1, col = rgb(rgbImage[, 3:5]), asp = 1, pch = "."))
 
# Cluster in color space:
kColors <- 5 # Number of palette colors
kMeans <- kmeans(rgbImage[, 3:5], centers = kColors)
 
zp1 <- qplot(factor(kMeans$cluster), geom = "bar",
fill = factor(kMeans$cluster))
zp1 <- zp1 + scale_fill_manual(values = rgb(kMeans$centers))
zp1
 
approximateColor <- kMeans$centers[kMeans$cluster, ]
with(rgbImage, plot(X2, X1, col = rgb(approximateColor), asp = 1, pch = "."))

Hi "dsparks" , very nice code, very useful to me.

I know you wrote this ages ago, but I am wondering, is there a simple way to get the name of the colours (perhaps just the two or three dominant colours)? My idea eventually would be measure the "distance" between two colours. But to start with, extracting the name would be enough, or a rgb code, or something that allows me to compare with other images.

Thanks in advance for your time ;)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Something went wrong with that request. Please try again.