Part 1.2 Car Nd Masking and Colouring a Region of the Image

May 9, 2017 • Python Machine-Learning Computer-Vision • 4 minutes to read • Edit

Masking regions

In this part of Car-ND, we will look at how to mask and colour the region.

Table of Contents

1 Basic Math

Before I show you the code, let’s understand some (very) basic mathematics that are necessary to understand for this part

1.1 One Degree Polynomial

A polynomial equation looks something like this (lets call it equation 1):

2 x 2 + 5 x + 1

2 Let’s Code

Let’s import all the necessary libraries

    import matplotlib.pyplot as plt
    import matplotlib.image as mpimg
    import numpy as np
    import sys

Now, let load in the image and get its shape and also while we are at it make a copy of the image. lets call the copies as color_select & line_image

    try:
        image = mpimg.imread('test.jpg')
    except FileNotFoundError as e:
        print(e)
        sys.exit(1)
    print('This image is: {}, with dimensions: {}'.format(type(image), image.shape))

    ysize = image.shape[0]
    xsize = image.shape[1]
    color_select = np.copy(image)
    line_image = np.copy(image)

Let’s initialize threshold values and a triangle coordinates.

    red_threshold = 200
    green_threshold = 200
    blue_threshold = 200
    rgb_threshold = [red_threshold, green_threshold, blue_threshold]

    # Init a triangle (this part is variable)
    left_bottom = [100, 525]  # [0, 539]
    right_bottom = [850, 550]  # [900, 539]
    apex = [460, 310]  # [475, 320]

Let’s get the threshold color image by taking RBG arrays of the image and getting all the values less then the predefined threshold values.

    color_thresholds = (image[:, :, 0] < rgb_threshold[0]) | (image[:, :, 1] < rgb_threshold[1]) | (image[:, :, 2] < rgb_threshold[2])

Using Numpy’s polyfit (more on that here) at 1 degree

    fit_left = np.polyfit((left_bottom[0], apex[0]), (left_bottom[1], apex[1]), 1)
    fit_right = np.polyfit((right_bottom[0], apex[0]), (right_bottom[1], apex[1]), 1)
    fit_bottom = np.polyfit((left_bottom[0], right_bottom[0]), (left_bottom[1], right_bottom[1]), 1)

Using Numpy’s Meshgrid (more on that here), let’s create a mesh

    xx, yy = np.meshgrid(np.arange(0, xsize), np.arange(0, ysize))

Getting the regional threshold is done by doing

    region_thresholds = (yy > (xx * fit_left[0] + fit_left[1])) & (yy > (xx * fit_right[0] + fit_right[1])) & (
    yy < (xx * fit_bottom[0] + fit_bottom[1]))

Before plotting the images, the final thing to do is

  • Masking the threshold region based on the triangle
  • And then masking the original image with the colored lines.
    # Mask color selection
    color_select[color_thresholds | ~region_thresholds] = [0, 0, 0]
    # Find where image is both colored right and in the region
    line_image[~color_thresholds & region_thresholds] = [255, 0, 0]

And finally, plot them

    # Show figures
    f = plt.figure()
    x = [left_bottom[0], right_bottom[0], apex[0], left_bottom[0]]
    y = [left_bottom[1], right_bottom[1], apex[1], left_bottom[1]]

    f.add_subplot(2, 3, 1)
    plt.plot(x, y, 'b--', lw=4)
    plt.imshow(image)  # Image superimposed with plotted lines
    plt.title("Step 1: Original image superimposed with plotted lines")

    f.add_subplot(2, 3, 2)
    plt.imshow(color_thresholds, cmap='gray')
    plt.title("Step 2: Get the threshold image")

    f.add_subplot(2, 3, 3)
    plt.imshow(region_thresholds, cmap='gray')
    plt.title("Step 3: Get the regional threshold")

    f.add_subplot(2, 3, 4)
    plt.imshow(color_select)  # bright color line with triangle
    plt.title("Step 4: Bright lines with triangle selected")

    f.add_subplot(2, 3, 5)
    plt.imshow(line_image)  # Masked image
    plt.title("Final: Original image with masked color")

    plt.show()

2.1 Complete Code

    import matplotlib.pyplot as plt
    import matplotlib.image as mpimg
    import numpy as np
    import sys

    try:
        image = mpimg.imread('test.jpg')
    except FileNotFoundError as e:
        print(e)
        sys.exit(1)
    print('This image is: {}, with dimensions: {}'.format(type(image), image.shape))

    ysize = image.shape[0]
    xsize = image.shape[1]
    color_select = np.copy(image)
    line_image = np.copy(image)

    red_threshold = 200
    green_threshold = 200
    blue_threshold = 200
    rgb_threshold = [red_threshold, green_threshold, blue_threshold]

    # Init a triangle (this part is variable)
    left_bottom = [100, 525]  # [0, 539]
    right_bottom = [850, 550]  # [900, 539]
    apex = [460, 310]  # [475, 320]

    # See https://docs.scipy.org/doc/numpy/reference/generated/numpy.polyfit.html for more info
    fit_left = np.polyfit((left_bottom[0], apex[0]), (left_bottom[1], apex[1]), 1)
    fit_right = np.polyfit((right_bottom[0], apex[0]), (right_bottom[1], apex[1]), 1)
    fit_bottom = np.polyfit((left_bottom[0], right_bottom[0]), (left_bottom[1], right_bottom[1]), 1)

    color_thresholds = (image[:, :, 0] < rgb_threshold[0]) | (image[:, :, 1] < rgb_threshold[1]) | (
    image[:, :, 2] < rgb_threshold[2])

    # See https://docs.scipy.org/doc/numpy/reference/generated/numpy.meshgrid.html for more info
    xx, yy = np.meshgrid(np.arange(0, xsize), np.arange(0, ysize))
    region_thresholds = (yy > (xx * fit_left[0] + fit_left[1])) & (yy > (xx * fit_right[0] + fit_right[1])) & (
    yy < (xx * fit_bottom[0] + fit_bottom[1]))

    # Mask color selection
    color_select[color_thresholds | ~region_thresholds] = [0, 0, 0]
    # Find where image is both colored right and in the region
    line_image[~color_thresholds & region_thresholds] = [255, 0, 0]

    # Show figures
    f = plt.figure()
    x = [left_bottom[0], right_bottom[0], apex[0], left_bottom[0]]
    y = [left_bottom[1], right_bottom[1], apex[1], left_bottom[1]]

    f.add_subplot(2, 3, 1)
    plt.plot(x, y, 'b--', lw=4)
    plt.imshow(image)  # Image superimposed with plotted lines
    plt.title("Step 1: Original image superimposed with plotted lines")

    f.add_subplot(2, 3, 2)
    plt.imshow(color_thresholds, cmap='gray')
    plt.title("Step 2: Get the threshold image")

    f.add_subplot(2, 3, 3)
    plt.imshow(region_thresholds, cmap='gray')
    plt.title("Step 3: Get the regional threshold")

    f.add_subplot(2, 3, 4)
    plt.imshow(color_select)  # bright color line with triangle
    plt.title("Step 4: Bright lines with triangle selected")

    f.add_subplot(2, 3, 5)
    plt.imshow(line_image)  # Masked image
    plt.title("Final: Original image with masked color")

    plt.show()

2.2 Output

Masking and region colouring:

Masking regions

Masking regions

3. What’s Next?

Canny edge detection, which looks something like:

Canny edges.




comments powered by Disqus