What is commonly used for features when classifying images with nearest neighbor algorithms?

If one wanted to perform the K-nearest-neighbors algorithm to do classification on im开发者_如何学Pythonages, how are features extracted from the images? What are the easiest, most effective methods?

---

Great tutorial on feature learning for image classification

---

It really depends on the exact problems.

For general problems, I'd always start with either some form of texture features (if you want image-level features) or local features (such as SURF).

The documentation for mahotas has a bit of an intro into this.

---

There are many features you can use... but as already said it depends on the case and what you get from your segmentations...

you can classify from:

color
rectangularity
circularity
area
diameter
circumference
...

these are just a few but should be easy to calculate once you segmented your objects...

Hope that helps...