What is it about?
This project investigates the impact of noise on train-captured images and how well different filtering techniques minimize it. A software application has been designed to simulate two common types of noise, namely Gaussian and salt-and-pepper, on sample images. Three filters—averaging, median, and Wiener—have been operated for different sizes of filters (3x3, 5x5, 7x7, 9x9) to observe their capabilities at eliminating noise. The results suggest that any filter can handle noise to some extent, but with an increase in size, a filter shows a trade-off. The higher the size of the filter, the more reduction in noise; however, it also blurs the image more. In general, the median filter is more efficient at keeping image details than the averaging filter. The Wiener filter copes quite well with Gaussian noise but works poorly with salt-and-pepper, especially at higher sizes. This work, therefore, has yielded some value in the development of image processing algorithms for train obstacle detection systems that operate under diverse weather conditions. Future work can also extend to a more realistic image dataset with a higher variety of noise types and combination of filters.
Featured Image
Photo by Pawel Czerwinski on Unsplash
Why is it important?
This project is important because train obstacle detection systems depend on clear and reliable images to identify hazards on or near the railway track. In real operating conditions, train-captured images can be affected by noise caused by weather, lighting, camera limitations, motion, or environmental interference. If this noise is not reduced properly, the system may fail to detect obstacles accurately, which can create safety risks. The study is also important because it compares different filtering techniques and shows that noise reduction is not simply about removing as much noise as possible. Larger filters may reduce more noise, but they can also blur important image details. This matters because details such as object edges, shapes, and boundaries are essential for detecting obstacles. By testing averaging, median, and Wiener filters on Gaussian and salt-and-pepper noise, the project helps identify which filters are more suitable for different noise conditions. For example, the median filter is useful for preserving details, while the Wiener filter performs better with Gaussian noise. These findings can support the development of more accurate and dependable image-processing algorithms for railway safety systems, especially under challenging weather and environmental conditions.
Perspectives
From my perspective, this project is important because image quality plays a major role in the reliability of train obstacle detection systems. Since train-captured images can be affected by noise due to weather conditions, poor lighting, camera limitations, or movement, it is necessary to study how different filters can reduce noise while preserving useful image details. This project helped me understand that filtering is not only about removing noise, but also about maintaining important visual features such as edges and object shapes. The results show that increasing the filter size can reduce more noise, but it can also cause more blurring. Therefore, choosing the right filter and filter size is very important. I also observed that the median filter performs better than the averaging filter in preserving image details, especially when dealing with salt-and-pepper noise. The Wiener filter gives better results with Gaussian noise, but it is less effective for salt-and-pepper noise, particularly when larger filter sizes are used. Overall, this project provides useful insight into how image processing techniques can improve the performance of train obstacle detection systems. It also shows that future research should use more realistic train image datasets, include different types of noise, and explore combinations of filters to achieve better results under real-world conditions.
HTMA Haeder Talib Alahmar
Al Furat Al Awsat Technical University
Read the Original
This page is a summary of: Study and evaluate the performance of image processing algorithms to remove noise in images in metro systems, January 2025, American Institute of Physics,
DOI: 10.1063/5.0286315.
You can read the full text:
Contributors
The following have contributed to this page
