Sotirios Diamantas
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sotirios Diamantas has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Bryant Wyatt
Department of Mathematics, College of Science and Mathematics, Tarleton State University, The Texas A&M University System, Stephenville, TX 76402, USA
2 shared publications - Ethan Reaves
Department of Computer Science and Electrical Engineering, Mayfield College of Engineering, Tarleton State University, The Texas A&M University System, Stephenville, TX 76402, USA
2 shared publications - Ethan Welborn
Department of Computer Science and Electrical Engineering, Tarleton State University, The Texas A&M University System, Stephenville, USA
2 shared publications Anastasios Oikonomidis
Centre for Risk Research, School of Management, University of Southampton, Southampton, SO17 1BJ, UK
1 shared publication- Joseph Meier
Department of Computer Science and Electrical Engineering, Tarleton State University, The Texas A&M University System, Stephenville, USA
1 shared publication Kostas Alexis
University of Nevada, Reno
1 shared publication
Latest Publications
- Open Access
- 22 December 2025
- Published by MDPI AG in Mathematics
- Vol. 14 (1) , 43
A Parallel Algorithm for Background Subtraction: Modeling Lognormal Pixel Intensity Distributions on GPUs
Background subtraction is a core preprocessing step for video analytics, enabling downstream tasks such as detection, tracking, and scene understanding in applications ranging from surveillance to transportation. However, real-time deployment remains challenging when illumination changes, shadows, and dynamic backgrounds produce heavy-tailed pixel variations that are difficult to capture with simple Gaussian assumptions. In this work, we propose a fully parallel GPU implementation of a per-pixel background model that represents temporal pixel deviations with lognormal distributions. During a short training phase, a circular buffer of n frames (as small as ) is used to estimate, for every pixel, robust log-domain parameters . During testing, each incoming frame is compared against a robust reference (per-pixel median), and a lognormal cumulative density function yields a probabilistic foreground score that is thresholded to produce a binary mask. We evaluate the method on multiple videos under varying illumination and motion conditions and compare qualitatively with widely used mixture of Gaussians baselines (MOG and MOG2). Our method achieves, on average, 87 fps with a buffer size of 10, and reaches about 188 fps with a buffer size of 3, on an NVIDIA 3080 Ti. Finally, we discuss the accuracy–latency trade-off with larger buffers.- 22 December 2025
- Published by MDPI AG in Mathematics
- Vol. 14 (1) , 43
Open AccessA Parallel Algorithm for Background Subtraction: Modeling Lognormal Pixel Intensity Distributions on GPUs
Background subtraction is a core preprocessing step for video analytics, enabling downstream tasks such as detection, tracking, and scene understanding in applications ranging from surveillance to transportation. However, real-time deployment remains challenging when illumination changes, shadows, and dynamic backgrounds produce heavy-tailed pixel variations that are difficult to capture with simple Gaussian assumptions. In this work, we propose a fully parallel GPU implementation of a per-pixel background model that represents temporal pixel deviations with lognormal distributions. During a short training phase, a circular buffer of n frames (as small as ) is used to estimate, for every pixel, robust log-domain parameters . During testing, each incoming frame is compared against a robust reference (per-pixel median), and a lognormal cumulative density function yields a probabilistic foreground score that is thresholded to produce a binary mask. We evaluate the method on multiple videos under varying illumination and motion conditions and compare qualitatively with widely used mixture of Gaussians baselines (MOG and MOG2). Our method achieves, on average, 87 fps with a buffer size of 10, and reaches about 188 fps with a buffer size of 3, on an NVIDIA 3080 Ti. Finally, we discuss the accuracy–latency trade-off with larger buffers. - Open Access
- 5 June 2025
- Published by MDPI AG in Preprints.org
A Parallel Algorithm for Background Subtraction: Modeling Lognormal Pixel Intensity Distributions on GPUs
In this research we address the problem of background subtraction using a parallel algorithm implemented and tested on GPU processors and compare our findings with the state-of-the-art. Our results demonstrate that our algorithm achieves a high number of processed frames per second while maintaining output quality equivalent to leading-edge algorithms. Background subtraction has applications in many areas, ranging from automotive and tool inspection to object detection, recognition, and tracking. While current algorithms commonly rely on mixture of Gaussian models, our algorithm is trained with a small set of sample frames, modeling pixels with lognormal distributions. During the testing phase, we compare pixel intensities with the lognormally distributed values from training with the view to infer a probabilistic score, which classifies each pixel as either foreground or background.
- 5 June 2025
- Published by MDPI AG in Preprints.org
Open AccessA Parallel Algorithm for Background Subtraction: Modeling Lognormal Pixel Intensity Distributions on GPUs
In this research we address the problem of background subtraction using a parallel algorithm implemented and tested on GPU processors and compare our findings with the state-of-the-art. Our results demonstrate that our algorithm achieves a high number of processed frames per second while maintaining output quality equivalent to leading-edge algorithms. Background subtraction has applications in many areas, ranging from automotive and tool inspection to object detection, recognition, and tracking. While current algorithms commonly rely on mixture of Gaussian models, our algorithm is trained with a small set of sample frames, modeling pixels with lognormal distributions. During the testing phase, we compare pixel intensities with the lognormally distributed values from training with the view to infer a probabilistic score, which classifies each pixel as either foreground or background.
- 3 December 2024
- Published by Springer Nature in Machine Vision and Applications
- Vol. 36 (1) , 1-12
Pothole segmentation and area estimation with thermal imaging using deep neural networks and unmanned aerial vehicles
- 3 December 2024
- Published by Springer Nature in Machine Vision and Applications
- Vol. 36 (1) , 1-12
Pothole segmentation and area estimation with thermal imaging using deep neural networks and unmanned aerial vehicles
- 3 December 2023
- Published by Springer Nature
- p. 370-384
Pothole Segmentation and Area Estimation with Deep Neural Networks and Unmanned Aerial Vehicles
- 3 December 2023
- Published by Springer Nature
- p. 370-384
Pothole Segmentation and Area Estimation with Deep Neural Networks and Unmanned Aerial Vehicles
- 7 December 2020
- Published by Springer Nature
- p. 398-409
Optical Flow Based Background Subtraction with a Moving Camera: Application to Autonomous Driving
- 7 December 2020
- Published by Springer Nature
- p. 398-409
Optical Flow Based Background Subtraction with a Moving Camera: Application to Autonomous Driving
- 24 April 2019
- Published by Springer Nature
- p. 551-563
Towards Resolving the Kidnapped Robot Problem: Topological Localization from Crowdsourcing and Georeferenced Images
- 24 April 2019
- Published by Springer Nature
- p. 551-563
Towards Resolving the Kidnapped Robot Problem: Topological Localization from Crowdsourcing and Georeferenced Images
- 1 December 2010
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 7-12
Depth computation using optical flow and least squares
Depth computation in robotics is an important step towards providing robust and accurate navigation capabilities to a mobile robot. In this paper we examine the problem of depth estimation with the view to be used in parsimonious systems where fast and accurate measurements are critical. For this purpose we have combined two methods, namely optical flow and least squares in order to infer depth estimates between a robot and a landmark. In the optical flow method the variation of the optical flow vector at varying distances and velocities is observed. In the least squares method snapshots of a landmark are taken from different robot positions. The results of the two methods show that there is a significant increase in depth estimation accuracy by combining optical flow and least squares.
- 1 December 2010
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 7-12
Depth computation using optical flow and least squares
Depth computation in robotics is an important step towards providing robust and accurate navigation capabilities to a mobile robot. In this paper we examine the problem of depth estimation with the view to be used in parsimonious systems where fast and accurate measurements are critical. For this purpose we have combined two methods, namely optical flow and least squares in order to infer depth estimates between a robot and a landmark. In the optical flow method the variation of the optical flow vector at varying distances and velocities is observed. In the least squares method snapshots of a landmark are taken from different robot positions. The results of the two methods show that there is a significant increase in depth estimation accuracy by combining optical flow and least squares.