CSCI 3240U --- Computer Vision I

Lab 8 --- Homography and Image Stitching

Faisal Z. Qureshi
Faculty of Science, Ontario Tech University
Oshawa ON Canada
http://vclab.science.ontariotechu.ca

Fall 2026

Copyright information

© Faisal Qureshi

License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Outline

Consider the following image pair

You are asked to pick at least four correspondences in both images. E.g., you can pick four points $(x^l_i,y^l_i)$ in the left image and then you can pick the corresponding locations $(x^r_i,y^r_i)$ in the right image in the same order. Here $i > 3$. Given these correspondences, estimate the homography matrix between the two set of images, and use this matrix to stitch the two images into a single image.

Picking locations in an image

Use the following code to pick locations on an image as seen below

Modify the code as needed to record the $(x,y)$ locations.

%matplotlib tk
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import cv2 as cv
import numpy as np

class PointPicker:
    def __init__(self, imgplot, img, color=(255,0,0), radius=20, thickness=20):
        self.imgplot = imgplot
        self.img = img
        self.color = color
        self.radius = radius
        self.thickness = thickness
        self.cid = imgplot.figure.canvas.mpl_connect('button_press_event', self)
    
    def __call__(self, event):
        if event.inaxes != self.imgplot.axes: 
            return
        ix = event.xdata
        iy = event.ydata
        print(f'x={ix}, y={iy}')
        self.img = cv.circle(self.img, (int(ix), int(iy)), self.radius, self.color, self.thickness)
        
        imgplot.set_array(self.img)
        self.imgplot.figure.canvas.draw()

img=cv.imread('1-left.jpeg')        
        
fig = plt.figure(figsize=(10,10))
ax = fig.add_subplot(111)
imgplot = plt.imshow(img)
point_picker = PointPicker(imgplot, img)
plt.show()

Task

• Load left image (1-left.jpeg) and pick at least four points in order.
• Load right image (1-right.jpeg) and pick the corresponding locations.
• Use the correspondences to estimate homography.
• Use the homography to stitch the images together.

Some outdoor photography

• The above image exhibits strong parallex. You may consider going for a walk and taking an image pair of your favorite neighbourhood landscape. Try not to translate the camera, just rotate in place.

Useful information

• Least squares fitting
• RANSAC
• Homography
• Image sampling

Bonus

• Try stitching on different image pairs.
• Can you stich more than one images to create a panorama?

Submission

Include code and stiched image single jupyter notebook. Submit via canvas.