Skip to content

Instantly share code, notes, and snippets.

@dsparks
Last active June 27, 2019 00:10
  • Star 10 You must be signed in to star a gist
  • Fork 4 You must be signed in to fork a gist
Star You must be signed in to star a gist
Save dsparks/3980277 to your computer and use it in GitHub Desktop.
Image Manipulation, Part 2
doInstall <- TRUE # Change to FALSE if you don't want packages installed.
toInstall <- c("jpeg", "reshape2", "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 <- readJPEG("tempPicture.jpg")
longImage <- melt(readImage)
rgbImage <- reshape(longImage, timevar = "Var3",
idvar = c("Var1", "Var2"), direction = "wide")
rgbImage$Var1 <- -rgbImage$Var1
return(rgbImage)
}
##########
# Part 2 # Identifying "dominant" colors with k-means
##########
rgbImage <- imageLoader("https://pbs.twimg.com/media/CGA2pXkWMAEl2Pb.jpg:large") # Pick one, or use your own URL.
with(rgbImage, plot(Var2, Var1, col = rgb(rgbImage[, 3:5]), asp = 1, pch = "."))
# Cluster in color space:
kColors <- 6 # 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, ]
qplot(data = rgbImage, x = Var2, y = Var1, fill = rgb(approximateColor), geom = "tile") +
coord_equal() + scale_fill_identity(guide = "none")
@jrzaurin
Copy link

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 ;)

@jbkunst
Copy link

jbkunst commented Mar 6, 2015

Hi dsparks,

Thanks for the code. I made a shiny app using your code, and adding others minimal features, obviously I put a link to your gist for people can see you profile and code.

The url is https://jbkunst.shinyapps.io/kmeans-images/

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment