regionGrowing.py

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import colors


def onClick(event):
    """Call back function to run the region growing algorithm

    Args:
        event (MouseClick): the mouse click event
    """
    # Get position of selected point
    x, y = int(event.xdata), int(event.ydata)
    # Start growing from this seed with specified threshold
    outRegion = regionGrowing(valChannel, np.array([y, x]), threshold)
    # Show segmente image
    showSegmentedImage(outRegion)


def showSegmentedImage(region):
    """Show the result of region growing algorithm

    Args:
        region (np.array): the segmented region
    """
    # Show only segmented regions
    segmentedImage = np.copy(myImage)
    segmentedImage[..., 0] *= region.astype('uint8')
    segmentedImage[..., 1] *= region.astype('uint8')
    segmentedImage[..., 2] *= region.astype('uint8')
    plt.figure()
    plt.imshow(segmentedImage)
    plt.show()


def regionGrowing(image, seed, threshold):
    """Region based image segmentation algorithm

    Args:
        image (np.array): gray image to be
        seed (tuple): starting point
        threshold (int): threshold for joining the region

    Returns:
        np.array: the segmented region
    """
    # Initilize points stack
    points = []
    # Initialize output image
    outImage = np.zeros(image.shape)
    # Add seed point to the list
    points.append(seed)
    # Add seed point to segmented region in output image
    outImage[seed[0], seed[1]] = 1
    # Loop till no point in the stack
    while len(points):
        # Pop a point
        p = points.pop()
        # Get its four neighbors
        neighbors = np.array([[p[0]-1, p[1]],
                              [p[0]+1, p[1]],
                              [p[0], p[1]-1],
                              [p[0], p[1]+1]])
        # Try for boundary condition
        try:
            # For all neighbors
            for i in range(4):
                # Check that if its already added to region or not
                if not outImage[neighbors[i, 0], neighbors[i, 1]]:
                    # Checking similarity based on intensity value
                    if np.abs(image[neighbors[i, 0], neighbors[i, 1]] -
                              image[p[0], p[1]]) <= threshold:
                        # Add point to region
                        outImage[neighbors[i, 0], neighbors[i, 1]] = 1
                        # Push it to the stack
                        points.append(np.array([neighbors[i, 0],
                                                neighbors[i, 1]]))
        except:
            continue

    return outImage


if __name__ == '__main__':
    # Read the image
    myImage = plt.imread('images/MRIbrain3.jpg')
    hsvImage = colors.rgb_to_hsv(myImage)
    # Get Intensity channel
    valChannel = hsvImage[..., 2]
    # Set similarity threshold
    threshold = 5
    # Show the original Image
    fig = plt.figure()
    plt.imshow(myImage)
    plt.set_cmap('gray')
    # Enable mouse event
    fig.canvas.mpl_connect('button_press_event', onClick)
    plt.show()