Canny Edge Detector

The Canny Edge Detector is an algorithm used to perform edge detection in image processing. This repository contains an implementation in Python.

Usage

Run python src/main.py <image-name> <low-threshold?> <high-threshold?> to execute the algorithm, where:

• <image-name> is the input image (that must be contained in input/ directory);
• <low-threshold?> and <high-threshold?> are the low and high thresholds to be used in hysteresis threshold step. These parameters are optional and their default values in the implementation are 20 and 40, respectively. Note: if only the low threshold is provided, the high threshold will be the double of it.

The output images will be stored in output/ directory, which are:

• _output-step-1.png: grayscale smoothed/blurred image, obtained using Gaussian kernel;
• _output-step-2.png: horizontal gradient of the image, obtained using Sobel kernel;
• _output-step-3.png: vertical gradient of the image, obtained using Sobel kernel;
• _output-step-4.png: gradient magnitude of the image, obtained using Pythagorean Theorem;
• _output-step-5.png: non-maximum suppressed image;
• output-detected-edges.png: the actual result image, which is obtained from the application of the hysteresis thresholding on the non-maximum suppressed image.

Examples

Flower Image

1. Input:
   
2. Smoothed:
   
3. Horizontal gradient:
   
4. Vertical gradient:
   
5. Gradient magnitude:
   
6. Non-maximum suppressed:
   
7. [Result] Hysteresis thresholding:
   

Bird Image

1. Before:
   
2. After:
   

Butterfly Image

1. Before:
   
2. After:
   

Egypt Image

1. Before:
   
2. After:
   

Python Notebook

A step-by-step explanation (in portuguese) of the project code can be read in a python notebook file.

License

This project is licensed under the MIT License.