Comparison of Image Feature Detection Algorithms

The key of image processing is to extract feature points and feature vectors by appropriate methods. In order to analyze the application effect of common feature extraction methods in different scenarios, this paper adopts SIFT, BRISK, ORB, KAZE and AKAZE methods to rotate, adjust brightness, blur a...

Full description

Saved in:
Bibliographic Details
Published inInternational Conference on Dependable Systems and Their Applications (Online) pp. 723 - 731
Main Authors Xu, Fan, Liu, Xia, Cui, Yanli, Yan, Mingdie, Lai, Zhongyuan
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2022
Subjects
AKAZE
Brightness
BRISK
Feature detection
Feature extraction
KAZE
ORB
Real-time systems
SIFT
Online AccessGet full text
ISSN2767-6684
DOI10.1109/DSA56465.2022.00103

Cover

More Information
Summary:The key of image processing is to extract feature points and feature vectors by appropriate methods. In order to analyze the application effect of common feature extraction methods in different scenarios, this paper adopts SIFT, BRISK, ORB, KAZE and AKAZE methods to rotate, adjust brightness, blur and zoom images based on two data sets. The operating efficiency, the ratio of feature points before and after operation, the best matching rate and accuracy rate of feature points pair of the five methods are analyzed, and the suggestions for selecting reasonable feature detection methods for different application scenarios are given.
ISSN:2767-6684
DOI:10.1109/DSA56465.2022.00103