This is the most simple feature descriptor I can imagine. It just samples and integrates along radial lines as shown in the figure below and the sum is stored as an element in the feature vector. Nonetheless, it works very well!
This vector was the base for the RLF descriptor, which also includes sampling using a Gaussian mask. This descriptor samples on pixel centers and is therefore many times faster.
- An Efficient Descriptor based on Radial Line Integration for Fast non Invariant Matching and Registration of Microscopy Images. A. Hast, G. Kylberg, I-M. Sintorn. Advanced Concepts for Intelligent Vision Systems (ACIVS2017, Full Paper. pp. 723-734. 2017.
@inproceedings{Has17a, author = {A. Hast and G. Kylberg and I-M. Sintorn}, title = {An Efficient Descriptor based on Radial Line Integration for Fast non Invariant Matching and Registration of Microscopy Images}, booktitle = {Advanced Concepts for Intelligent Vision Systems (ACIVS2017}, pages = {723--734}, year = {2017}}
Anders Hast 