Show HN: Reorganize pixels of a painting into a smooth color palette using PCA | Hacker News Reader

Show HN: Reorganize pixels of a painting into a smooth color palette using PCA

github.com

95 pointsdontreact30 comments

Shameless plug/relevant background: I did this in order to make a flyer for a show I am playing with my band Congratulatinos in Oakland, CA on Dec. 17th Band: http://congratulatinos.bandcamp.com/ Event: https://www.facebook.com/events/661571907356009/

I opened a small PR - scipy was missing as a requirement (I'm guessing you already had it installed, and didn't stick it in requirements.txt)

thanks. right on all counts.

This looks interesting. I was wondering if you could explain a little more about what information about a picture this transform is exposing? I mean, if you look at just the output image, what kind of insight can you gain into the original input image?

For some paintings I think you can get a sense of what paints the painter used.

Just to clarify: you started with the pic of the 2 guys and created the fading effect used to color the writing with pca?

That's pretty awesome, I think the final effect is very pleasant to the eyes.

The script merely remaps the pixels in the input to new locations via PCA. I'm not 100% sure I know what you mean by "2 guys" or "color the writing".

The two guys on the page you just linked. The colored writing on the page you just linked.

Huh still confused somehow, sorry. This is the input https://github.com/ardila/paintingReorganize/blob/master/inp...

My script moves the pixels of that image around to get this output https://github.com/ardila/paintingReorganize/blob/master/out...

They are asking how the generated gradient from the script was incorporated into the final image of the band promo.

The flyer is in the facebook event not the band page. The cover photo for the album was a completely separate project :)

Ahhh. I tiled the image 4 times, remapped to polar coordinates, then did some color balancing.

Did you also experiment with a simple sorting algorithm and if so, how were the results different?

I think there's a bug in there that flipped a few columns - but I actually really like the output better that way: http://imgur.com/RErXOF6

That is weird. It decides to flip a column by summing up the vector distances between the pixels in that column vs. the pixels in the previous column. I'm wondering if that since there is so much gray in the painting, this causes a large accumulation of error in gray vs. slightly different gray which outweighs the orange vs gray.

I do know that euclidean distance in RGB space is not that great a metric for perceptual similarity. Perhaps everything should be done in YUV instead? I will try tonight. Thanks.

I recommend you to try the LAB color space and to use delta E 94 [1] for computing distances between colors in the color space

[1] https://en.wikipedia.org/wiki/Color_difference#CIE94

The colors in that image seem super bright compared to the original.

https://www.moma.org/collection/works/79971?locale=en

The output for that version of the painting looks similar. I just image searched for "kupka" and tried a few different results.

Result for your image: http://imgur.com/fp9i8BP

Do you think I should update the repo with this version?

I do, only because it's from the institution that actually owns the real painting. Museums try pretty hard to get colors right on the digital images that they put online. And since the project is about color, it seems like an important detail. Thanks for being open to my nitpicking. Very cool project.

Made somewhat similar project by organizing pixels in three dimensional space of HSL (hue saturation lightness) color model: http://www.pixelbox.com/colorclouds/

Note that this reduces identical pixels into one point in 3D space. The key idea behind the experiment was to learn more about the palettes used by painters.

I love pixel sorting! I had seen it done by rows or cols before, but never the whole thing, and the result is beautiful.

I'm curious where the noise is coming from, as some pixels appear to be out of place.

Yes: it seems to be for colors where they are similar to two or more other clusters of pixels. I imagine that these are just really hard to place without seeming too out of place? Also probably has to do with the column by column "greedy" approach that isn't really global.

Probably some form of dithering, I'd guess. I'm curious myself.

In one Iain M Banks story, it is mentioned that a character was ordered not to destroy a library, so he had all the paintings given the treatment provided by this software and all the books cut up and arranged so that individual letters were in alphabetical order.

My first thought was "paintings don't have pixels"; but colour is quantised I guess. I wonder how many pixels you need to get the same graduation between color tones as the original image, presumably enough to provide atom level resolution.

I believe the term you want is "discretized", not "quantized". They're quantized too, but that's about there being a finite number of shades rather than a finite number of subdivisions of the image.

As a graphic designer turned data scientist, that is the coolest poster art backstory I have ever heard.

Are there any more examples?

I just put up 3 more. Starry night turned out very nicely. Anything you'd like to see in particular?

Reply on news.ycombinator.com