Computer Vision and Image Processing with OpenCV

OpenCV ‘Open Source Computer Vision Library’ is an open-source library that includes several hundreds of computer vision algorithms. Using OpenCV, you could pretty much do every Computer Vision task that could ever be imagined!

Real-life problems require you to utilize multiple building blocks together in order to achieve the desired outcome. So, you just have to understand what modules and functions are needed to get what you want!

In the next few minutes, we shall quickly dive into what OpenCV can do out of the box!

© Tanvi Penumudy 2020 — Rose Painting: Before and After Enhancement with OpenCV

Content Covered

The following topics have been precisely covered in the due course of the article —

1. Installation and Importing
2. Getting Started
3. Edge Detection
4. Feature Extraction
5. Feature Matching
6. Transformations
7. Face Detection
8. Working with Videos
9. Additional Resources and References

Installation and Importing

Installation

Can be done on your local machine via Python command prompt

pip install opencv-python

Refer opencv-python· PyPI for troubleshooting

Importing OpenCV

import cv2 # OpenCV-Python

It’s as simple as that!

Getting Started

Open/display an image

Modify pixels & ROI

• You can access/modify a single-pixel or ROI (Region of Interest) using Numpy indexing.
• Just like matrix indexing, img[a, b] refer to a-th row and b-th column.

Code Implementation: https://github.com/tanvipenumudy/Winter-Internship-Internity/blob/main/Day%2005/Day-5%20Notebook-1%20(OpenCV%20Basics).ipynb

Edge Detection

We shall now detect edges using Sobel operator cv2.Sobel() and Canny edge detectorcv2.Canny()

Edge Detection with Sobel Operator

dst = cv2.Sobel(src, ddepth, dx, dy, ksize=3, scale=1.0)

• src: input image
• ddepth: output image depth
• dx: order of the derivative x
• dy: order of the derivative y
• ksize: size of the extended Sobel kernel; it must be 1, 3, 5, or 7
• scale: optional scale factor for the computed derivative values

Canny Edge Detection

edges = cv2.Canny(image, threshold1, threshold2, apatureSize=3, L2gradient=False)

• image: 8-bit grayscale input image
• threshold1/threshold2: thresholds for the hysteresis procedure
• apertureSize: aperture size for the Sobel() operator
• L2gradient: A flag. True to use L2, else L1.

Code Implementation: https://github.com/tanvipenumudy/Winter-Internship-Internity/blob/main/Day%2005/Day-5%20Notebook-2%20(Edge%20Detection).ipynb

Feature Extraction

We shall now see how to extract SIFT (Scale-Invariant Feature Transform) and match SIFT features of two images with OpenCV-Python.

Extract SIFT features from an image

You need to initially convert the image into a greyscale image

Feature Matching

SIFT Feature Matching

Displaying Matches

Code Implementation: https://github.com/tanvipenumudy/Winter-Internship-Internity/blob/main/Day%2005/Day-5%20Notebook-3%20(Feature%20Extraction%20%26%20Matching).ipynb

Transformations

Let’s now see how to do basic operations on images such as — Scaling, Translation, Rotation, Affine Transformation and Homography (Perspective Transformation)

Scaling

Translation

Rotation

Affine Transformation

Homography

Code Implementation: https://github.com/tanvipenumudy/Winter-Internship-Internity/blob/main/Day%2005/Day-5%20Notebook-4%20(Transformations).ipynb

Face Detection

Face detection using ‘Haarcascades’ is a machine learning-based approach where a cascade function is trained with a set of input data. OpenCV already comprises of a good number of in-built classifiers for face, eyes, smiles, etc.

Code Implementation: https://github.com/tanvipenumudy/Winter-Internship-Internity/blob/main/Day%2005/Day-5%20Notebook-5%20(Face%20Detection).ipynb

Working with Videos

OpenCV library can be used to perform multiple operations on videos, apart from images which we’ve seen in the aforementioned sections.

Capture Video from Camera

To capture a video, we need to create a VideoCapture object. VideoCapture has the device index or the name of a video file. Device index is just the number to specify which camera. If we pass 0 then it is for the first camera, 1 for the second camera so on. We capture the video frame by frame.

cv2.VideoCapture(0) # Means first camera or webcam
cv2.VideoCapture(1) # Means second camera or webcam
cv2.VideoCapture("file name.mp4") # Means video file

Playing Video from file

It is same as capturing from Camera, just change camera index with the video file name. Also while displaying the frame, use appropriate time for cv2.waitKey().

Saving a Video

So we capture a video, process it frame-by-frame and we want to save that video. For images, it is very simple, just use cv2.imwrite().

Here a little more work is required!

You can also access some of the features of this video using cap.get(propId) method where propId is a number from 0 to 18. Each number denotes a property of the video and full details can be seen here Property Identifier.

That’s all for today! I know that’s a lot to take in at once! But you made it until the end! Kudos on that!

Also, do not forget to go through — OpenCV Documentation

Additional Resources and References

OpenCV has been around for a while and there are a lot of other good resources if you’re still interested in getting the most out of this library.

• https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_gui/py_video_display/py_video_display.html
• https://github.com/dalgu90/opencv-tutorial
• https://towardsdatascience.com/computer-vision-for-beginners-part-1-7cca775f58ef

For the complete implementation, do check out my GitHub Repository —

tanvipenumudy/Winter-Internship-Internity

GitHub Gists featured in this article —

• https://gist.github.com/tanvipenumudy/3e70b917371c521a9a72f267007eff58
• https://gist.github.com/tanvipenumudy/bfb60186cfa9b61c865fff8d9ed66490
• https://gist.github.com/tanvipenumudy/81f7fcf0407557bb6f377c64cdd1f80a