The form of these types of edge detection filters serve as a major inspiration for the use of learned convolutional filters in machine learning, since people found that the edges of objects serve as a useful feature for classifying them.
tristanburke
To add on to @keirp above, this method of edge detection can also be used for styling an image or render. For example, in the final project I'm working on, we are implementing cel-shading. In this style, objects are given a black outline to appear more comic-like. This edge detection method is sometimes used to identify the correct lines to blacken.
julialuo
If you're interested in edge detection algorithms, CS 280 goes more in depth and has a long homework question exploring different algorithms. Some examples are Gaussian derivative filters (convolving a gaussian with something like the Sobel matrices on this slide) to get smoother lines, and using oriented Gaussian derivative filters to catch different orientations of lines.
shivamparikh
Edge detection is also huge on machine learning topics for classification, as mentioned above, but something else that's cool is that it can be used for image alignment in panoramas by removing certain levels of noise from normalized cross-correlation computations for feature matching.
CptTeddy
Adding to the above discussion, besides creating interesting filters for images, edge detection can also be used for aligning images with the same edge information but different color information. An amazing historic use case implemented in CS194-26 can be found here: http://inst.eecs.berkeley.edu/~cs194-26/fa18/hw/proj1/.
tyleryath
This video gives a great overview of edge detection and sobel operators: https://www.youtube.com/watch?v=uihBwtPIBxM
The form of these types of edge detection filters serve as a major inspiration for the use of learned convolutional filters in machine learning, since people found that the edges of objects serve as a useful feature for classifying them.
To add on to @keirp above, this method of edge detection can also be used for styling an image or render. For example, in the final project I'm working on, we are implementing cel-shading. In this style, objects are given a black outline to appear more comic-like. This edge detection method is sometimes used to identify the correct lines to blacken.
If you're interested in edge detection algorithms, CS 280 goes more in depth and has a long homework question exploring different algorithms. Some examples are Gaussian derivative filters (convolving a gaussian with something like the Sobel matrices on this slide) to get smoother lines, and using oriented Gaussian derivative filters to catch different orientations of lines.
Edge detection is also huge on machine learning topics for classification, as mentioned above, but something else that's cool is that it can be used for image alignment in panoramas by removing certain levels of noise from normalized cross-correlation computations for feature matching.
Adding to the above discussion, besides creating interesting filters for images, edge detection can also be used for aligning images with the same edge information but different color information. An amazing historic use case implemented in CS194-26 can be found here: http://inst.eecs.berkeley.edu/~cs194-26/fa18/hw/proj1/.
This video gives a great overview of edge detection and sobel operators: https://www.youtube.com/watch?v=uihBwtPIBxM