Line-Like Object Detection in Corrupted Images by Averaging of Second Order Vector Fields

2021 IEEE 9th Workshop on Advances in Information, Electronic and Electrical Engineering (AIEEE 2021) 2021
Mihails Pudžs, Artūrs Āboltiņš, Sandis Migla, Oskars Šēlis, Pauls Ēriks Šics, Dans Laksis, Anna Litviņenko

In this paper the possibility of improving signal-to-noise ratio (SNR) of detected line-like objects (LLOs) by summing together second order vector fields (2Φs) acquired by processing the line images with added noise or other random disturbances are investigated. A 2Φ can be obtained by processing the input image with a LLO detection filter. The resulting vector field can be used for the extraction of the image features and enables detection of more complex objects. When adding many 2Φs together, persistent LLOs will be amplified, whereas randomly occurring LLOs will tend to mutually attenuate. Images with added Gaussian white noise, Salt and Pepper noise, line and LLO disturbances are examined and SNR values examined statistically. The resulting SNR values are compared with the classical image averaging techniques.

Atslēgas vārdi
computer vision , object detection , statistical analysis , image processing , filtering , signal processing , computing , noise
DOI
10.1109/AIEEE54188.2021.9670270
Hipersaite
https://ieeexplore.ieee.org/document/9670270

Pudžs, M., Āboltiņš, A., Migla, S., Šēlis, O., Šics, P., Laksis, D., Litviņenko, A. Line-Like Object Detection in Corrupted Images by Averaging of Second Order Vector Fields. No: 2021 IEEE 9th Workshop on Advances in Information, Electronic and Electrical Engineering (AIEEE 2021), Latvija, Rīga, 25.-26. novembris, 2021. Piscataway: IEEE, 2021, 1.-6.lpp. ISBN 978-1-6654-6713-1. e-ISBN 978-1-6654-6712-4. ISSN 2689-7334. e-ISSN 2689-7342. Pieejams: doi:10.1109/AIEEE54188.2021.9670270

Publikācijas valoda
English (en)