Research on Real-Time Feature Extraction Algorithm for Image Processing

Abstract

In processing images, the detection and matching of feature points occupy an extremely critical position. Considering the actual application scenarios in image recognition and processing problems, this paper designs experiments on the traditional SIFT (Scale-invariant feature transform), SURF (Speeded Up Robust Features), ORB (Oriented FAST and Rotated BRIEF) three algorithms from the four aspects of algorithm execution speed, lighting conditions, rotation processing, and blurred pixels, and compares the experimental results. We have the following conclusions: the ORB algorithm has a faster execution speed, and the ratio of execution speed of SIFT, SURF, ORB algorithm is about 2:1:14. In terms of feature point matching, the advantage of SIFT algorithm is more prominent, and it can still show a higher matching rate in three complex situations. SURF is the most robust algorithm. Even in the simulation of various complex image transformation scenarios, it also performs better, SIFT comes next, and the ORB algorithm has the worst robustness.

Related Papers

• → A Comparative Evaluation of Well-known Feature Detectors and Descriptors(2014)70 cited
• → SIFT Based Monocular SLAM with GPU Accelerated(2018)5 cited
• → Image Matching Using SIFT, SURF, BRIEF and ORB: Performance Comparison for Distorted Images(2017)303 cited
• → Research on Real-Time Feature Extraction Algorithm for Image Processing(2021)