Intuitively Understanding Harris Corner Detector

comsci.blog

184 points·anilz··8 comments

I remember the eigenvector analysis of the original paper [1] wasn't terribly inaccessible. I think an alternative title for this blog post could be "Intuitively Understanding the Harris Corner Detector Optimizations".

[1] https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&d...

I had to write one for an interview many moons ago. It's a fun little exercise.

PSA : The Harris Corner detector, while interesting to understand if you like linear algebra, is not exactly what you'd call state of the art in the feature detection layer of computer vision.

Yes? What is the state of the art?

The most widely used algorithms for classical feature detection today are "whatever opencv implements"

In terms of tech that's advancing at the moment? ML techniques. https://co-tracker.github.io/ if you want to track individual points, https://github.com/matterport/Mask_RCNN and its descendents if you want to detect, say, the cover of a book.

I believe the SIFT algorithm is most commonly used. Harris struggles when features change in scale between images, whereas SIFT does not. Harris can be outfitted with a Laplacian pyramid to overcome the scale issue though.

which is funny since SIFT is basically just as ancient.

Actually, it is a part of the ORB algorithm and ORB is especially used a lot in visual SLAM applications.

Reply on news.ycombinator.com