Visit the wiki pages to find some additional documentation and instructions on how view an interactive verson of these notebooks using binder.

---

%matplotlib inline

import math
import numpy as np
import matplotlib.pyplot as plt


from skimage import data
from skimage.color import rgb2gray
from skimage import io
from skimage import transform
from skimage.feature import (match_descriptors, corner_harris,
                             corner_peaks, ORB, plot_matches)

Detecting feature points in an image using ORB

Here we use the ORB feature detector as an alternative to highlight the principle of matching feature points.

rgb_image = io.imread("../images/mickey-mouse-image.jpg")
image = rgb2gray(rgb_image)
tform = transform.AffineTransform(scale=(1.3, 1.1), rotation=0.5,
                                  translation=(0, -200))
image_t = transform.warp(image, tform)

ffig, ax = plt.subplots(figsize=(12, 12))
ax = plt.subplot(1,2,1)
plt.imshow(image, cmap='gray')
ax.set_title('original image', y=1.05, fontsize=18)

ax = plt.subplot(1,2,2)
plt.imshow(image_t, cmap='gray')
ax.set_title('original image', y=1.05, fontsize=18)

Text(0.5, 1.05, 'original image')

Apply the keypoint detector

NK = 50
descriptor_extractor = ORB(n_keypoints=NK)

descriptor_extractor.detect_and_extract(image)
keypoints1 = descriptor_extractor.keypoints
descriptors1 = descriptor_extractor.descriptors

descriptor_extractor.detect_and_extract(image_t)
keypoints2 = descriptor_extractor.keypoints
descriptors2 = descriptor_extractor.descriptors

matches = match_descriptors(descriptors1, descriptors2, cross_check=True)

ffig, ax = plt.subplots(figsize=(36, 36))
ax = plt.subplot(1,2,1)
plot_matches(ax, image, image_t, keypoints1, keypoints2, matches)
ax.axis('off')
ax.set_title("Original Image vs. Transformed Image")


plt.show()

Explore: Replace the cartoon image used where with some of the medical images that are being provided. Observe how well the feature points are matched. Vary the value of parameter NK.