satyarth / pixelsort Goto Github PK

This is a feature request to add "curvy" pixel sorting options.

For some examples, austinhein_1 on Reddit has some posts that show the effect here and here.

In theory this should be simple to do but there are many details that might be a bit subtle. There's also the question of how to specify the curves which I don't have a good idea of what people should expect.

Algorithmic Idea

However the curve is specified, walk along the curve in the original image, filling in an array of pixels, then pixel sort as normal, providing the same threshold and sorting option as before.

The subtlety comes in doing sub pixel sampling and knowing how to sample and realize the pixels in the destination image.

Maybe I'm overthinking it and forcing the curve to only pick a single pixel at any given point, ignoring aliasing effects, is reasonable.

Curve Specification

I can think of a few different methods:

• Algebraic: Provide an explicit equation, maybe in the form of a complex equation
• Flow Field: Provide a file that has "velocity" (v_x,v_y) positions at key points, potentially at a lower sample rate than the input picture, and then trace curves over that flow field. This can be used in conjunction with a mask to make sure only the desired pixels get sorted
• Normal Map: somehow use a normal map image, interpreting the (r,g,b) -> (x,y,z) normal in some way and considering it a vector/flow field (?)

There might be some subtlety here with curves needing to be non-constant width when sorting pixels, because it might be hard to find "stacked" curves that only need on pixel.

Anyway, I'm not exactly sure how to implement this but it would be a cool feature.

It appears that the manifest is missing at least one file necessary to build
from the sdist for version 1.0.1. You're in good company, about 5% of other
projects updated in the last year are also missing files.

+ /tmp/venv/bin/pip3 wheel --no-binary pixelsort -w /tmp/ext pixelsort==1.0.1
Looking in indexes: http://10.10.0.139:9191/root/pypi/+simple/
Collecting pixelsort==1.0.1
  Downloading http://10.10.0.139:9191/root/pypi/%2Bf/86e/c9bbddb54a6ba/pixelsort-1.0.1.tar.gz (6.2 kB)
    ERROR: Command errored out with exit status 1:
     command: /tmp/venv/bin/python3 -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'/tmp/pip-wheel-rdx7czpx/pixelsort/setup.py'"'"'; __file__='"'"'/tmp/pip-wheel-rdx7czpx/pixelsort/setup.py'"'"';f=getattr(tokenize, '"'"'open'"'"', open)(__file__);code=f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' egg_info --egg-base /tmp/pip-wheel-rdx7czpx/pixelsort/pip-egg-info
         cwd: /tmp/pip-wheel-rdx7czpx/pixelsort/
    Complete output (5 lines):
    Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "/tmp/pip-wheel-rdx7czpx/pixelsort/setup.py", line 6, in <module>
        with open('requirements.txt') as fh:
    FileNotFoundError: [Errno 2] No such file or directory: 'requirements.txt'
    ----------------------------------------
ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.

I'm getting SyntaxErrors when trying to run the examples with the latest build (953a38c):

$ python3 pixelsort.py examples/image.jpg -i random -c 20
Traceback (most recent call last):
  File "pixelsort.py", line 6, in <module>
    from argparams import parse_args, verify_args
  File "/home/abe/git/github/satyarth/pixelsort/argparams.py", line 88
    logging.info(f"Interval function: {args['interval_function']}")
                                                                 ^
SyntaxError: invalid syntax

It looks like spurious fs got inserted on various lines from 88 to 100.

Removing those extra characters and I can run the example files.

A pull request will follow unless you get to it before me.

Hi
I got an error

$ python36 pixelsort.py test.png 
Interval function:  threshold
Lower threshold:  0.25
Upper threshold:  0.8
Randomness:  0 %
Opening image...
Converting to RGBA...
Rotating image...
Getting data...
Getting pixels...
Determining intervals...
Defining intervals...
Traceback (most recent call last):
  File "pixelsort.py", line 56, in <module>
    main()
  File "pixelsort.py", line 31, in main
    intervals = argparams.interval_function(pixels, argparams)
  File "/home/clone/repos/pixelsort/interval.py", line 60, in threshold
    if util.lightness(pixels[y][x]) < args.bottom_threshold or util.lightness(pixels[y][x]) > args.upper_threshold:
  File "/home/clone/repos/pixelsort/util.py", line 11, in lightness
    return rgb_to_hsv(pixel[0], pixel[1], pixel[2])[2] / 255.0  # For backwards compatibility with python2
TypeError: 'int' object is not subscriptable

While trying to use the angle parameter, I've noticed that the pixels don't all stay together in the original dimensions during the sorting:

• Immediately after rotation:
  
• After sorted pixels are placed:
  
  This causes the result image, after it is rotated back, to have some blank pixels. The sorting seems to be chopping up the right side of the image in a patter, which if I had to guess has to do with the intervals spanning over into areas with no real pixels.

I get the following error when trying to use the mask on a rotated pixel sort image:

$ python3 pixelsort.py examples/image.jpg -i random -a 120 -c 20 -m examples/mask.png 
Traceback (most recent call last):
  File "pixelsort.py", line 81, in <module>
    main(parse_args())
  File "pixelsort.py", line 33, in main
    pixels = get_pixels(data, mask, input_img.size)
  File "pixelsort.py", line 62, in get_pixels
    if not (mask and mask[x, y] == constants.black_pixel):
IndexError: image index out of range

After adding the following line:

    mask = Image.open(args["mask"]).convert('RGBA').rotate(args["angle"], expand=True).load() if args["mask"] else None

I get the following image as I would expect:

I'll submit a PR with the fix unless you get to it before I do.

While experimenting with pixelsort, I found that some sorting methods (-s) produce very similar results. Intensity is very similar to lightness, but it is actually different' so it's not too bad. maximum is much worse.

On all images I've tried, maximum produced identical output to lightness (default). This is easiest to observe if you run the app with none interval method, because it always behaves the same:

python3 pixelsort.py image.jpg -o test.png -i none -s maximum
python3 pixelsort.py image.jpg -o test.png -i none

Digging into it, I read the Wikipedia article on HSV colorspace. "Lightness" sorting method actually returns the "value" of HSV colorspace. "maximum" sorting method of the app takes max(pixel[0], pixel[1], pixel[2]). But how is "lightness" defined in HSV ?

"In the HSV "hexcone" model, value is defined as the largest component of a color, our M above"
https://en.wikipedia.org/wiki/HSL_and_HSV#Lightness

What is "M" ?

M = max(R, G, B)
https://en.wikipedia.org/wiki/HSL_and_HSV#Hue_and_chroma

So, unless I'm missing something, maximum and lightness look identical to me because they are identical, to the point of having the same definition!

The README lists only 5 real interval functions, but a sixth interval function file_edges is defined in interval.py. Is this a discrepancy in the README or is this a vestigial function? This is probably more of a question for @satyarth.

Hello!
If i try to run the script i get the following:

pixelsort.py:43: SyntaxWarning: "is not" with a literal. Did you mean "!="?
  if args["angle"] is not 0:

It doesn't stop the script from doing its job, though
Python version: 3.8.0

For sake of testing this, I'm trying to sort this image:

Using this external interval file:

The operation:
λ python pixelsort.py "pclips/pclips.png" -of pclips -a 45 -f "pclips/mask.png" -i file

Returns the following:

Interval function:  file
Randomness:  0 %
Opening image...
Converting to RGBA...
Rotating image...
Getting data...
Getting pixels...
Determining intervals...
Cleaning up edges...
Traceback (most recent call last):
  File "pixelsort.py", line 56, in <module>
    main()
  File "pixelsort.py", line 31, in main
    intervals = argparams.interval_function(pixels, argparams)
  File "C:\Users\justi\Code\pixelsort\interval.py", line 117, in file_mask
    if file_pixels[y][x] == constants.black_pixel and file_pixels[y][x - 1] == constants.black_pixel:
IndexError: list index out of range

I have also tested this on couple other bitmap mask files -- is there something I'm missing? Black pixels too close to the edge?

When using your program with the option -a 90, the first column of pixels is always completely transparent. This is not observed when this argument is not present (i.e. horizontal mode).