how to find the pixel passed by a plot drawn with matplotlib

user_10311197 0
Viewed 557

I plot a function using the following code:

t = np.arange(0., 5., 0.2)
plt.plot(t, (t**2)+10*np.sin(t))
plt.axis('off')

I am wondering how to save the plot as a 0/1 array, where a pixel value is 1 if the plot passes it, otherwise it's 0.

A follow up question is, if I draw the plot with some line width, I want the pixel value to be 1 only if it is on the "central" line of the plot, otherwise it's 0. How should I do it? Thank you!

pixel matplotlib numpy image python
1 Answers

Converting a figure into an RGBA array can be done in many ways. The simplest is probably to save the file as a PNG and then load the file again with plt.imread or similar. If that seems to roundabout for you, you can use the plot2img that I am using below, which grabs the canvas and converts it to an array via an intermediate representation as a string buffer.

After that it is just a matter of thresholding the image and extracting the medial axis, which can be easily done using functions provided by scikit-image.

enter image description here

#!/usr/bin/env python
"""
https://stackoverflow.com/q/62014554/2912349
"""

import numpy as np
import matplotlib.pyplot as plt

from matplotlib.backends.backend_agg import FigureCanvasAgg

from skimage.color import rgb2gray
from skimage.filters import threshold_otsu
from skimage.morphology import medial_axis


def plot2img(fig, remove_margins=True):
    # https://stackoverflow.com/a/35362787/2912349
    # https://stackoverflow.com/a/54334430/2912349

    if remove_margins:
        fig.subplots_adjust(left=0, bottom=0, right=1, top=1, wspace=0, hspace=0)

    canvas = FigureCanvasAgg(fig)
    canvas.draw()
    img_as_string, (width, height) = canvas.print_to_buffer()
    return np.fromstring(img_as_string, dtype='uint8').reshape((height, width, 4))


if __name__ == '__main__':

    t = np.arange(0., 5., 0.2)
    y = (t**2)+10*np.sin(t)

    # plot in a large figure such that the resulting image has a high resolution
    fig, ax = plt.subplots(figsize=(20, 20))
    ax.plot(t, y)
    ax.axis('off')

    # convert figure to an RGBA array
    as_rgba = plot2img(fig)

    # close plot made with non-interactive Agg backend so that we can open the other later
    plt.close('all')

    # threshold the image
    as_grayscale = rgb2gray(as_rgba)
    threshold = threshold_otsu(as_grayscale)
    as_bool = as_grayscale < threshold

    # find midline
    midline = medial_axis(as_bool)

    # plot results
    fig, (ax1, ax2) = plt.subplots(1, 2)
    ax1.imshow(as_bool, cmap='gray_r')
    ax2.imshow(midline, cmap='gray_r')
    plt.show()
user_29123490
Related