in most of the research articles, they have used Euclidean distance approach to classify images, if we take the moment of an image as a feature.
Feed forward neural network which is subdivision of Back Propagation Neural network is the best for pixel averaging feature.
Hi Vishal,
Euclidean Distance (ED) measure is a very well known "distance measure" that can be used in image analysis, however, whatever can be applied in statistics, it can be, generally applied in imaging. There are several ( numerous) "Distance Measures" (DM) which you can decide how to use. They are applicable when you are familiar with the type of the probability distribution you are studying. They also can be applied for solving problems of non-linearity. My guess is that ED is pretty straightforward DM.
See an example of several DMs. If you do a bit of research you may find some more complex DMs.
This, I hope resolves your question. You can also use the arithmetic average, log average, variance, kurtosis or any local (group of pixel-based) measures (k-means) to run classification in imaging.
https://stat.ethz.ch/education/semesters/ss2012/ams/slides/v4.2.pdf
I completely agree with Vera Van Raad, I made use of Simple distance Classifier when I did pixel-wise operation and my performance did not suffer performance inadequacy. So I suggest you make use of any simple distance classifier like Euclidean distance classifier.
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification