Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2026
- Goals
- Record
- 9951390b-a597-4796-9ef2-8d67458f56cc
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted a study comparing image enhancement methods for computer vision-based drowsiness detection in low-light conditions. The activity involved Subowo, Moh Hadi; Andono, Pulung Nurtantio; Shidik, Guruh Fajar; and Santoso, Heru Agus as participants. The study addressed the challenges of using near-infrared sensors in driver monitoring systems, where appropriate image enhancement is important for detection accuracy.
The evaluation systematically examined five methods: Histogram Equalization, Adaptive Histogram Equalization, Contrast-Limited Adaptive Histogram Equalization, Brightness Preserving Dynamic Histogram Equalization, and Multi-Scale Retinex with Color Restoration. The assessment used 4,272 frames from the University of Liège Multimodality Drowsiness Database. Four no-reference metrics were applied: Natural Image Quality Evaluator, Perception-based Image Quality Evaluator, Shannon Entropy, and Lightness Order Error. Additional validation measured the face detection rate using MediaPipe.
The findings identified Contrast-Limited Adaptive Histogram Equalization as achieving an optimal balance. It obtained an NIQE of 4.61, described as the best natural quality, a face detection rate of 97.9%, and an LOE of 0.058, indicating superior structural preservation. Multi-Scale Retinex with Color Restoration produced the highest entropy, at 6.58, but recorded the lowest face detection rate, at 75.6%. This result indicated structural distortion in the near-infrared context. Statistical validation using the Wilcoxon signed-rank test and the Friedman test confirmed the significance of the findings, with p < 0.05.
In relation to Sustainable Development Goal 3, the activity contributes evidence from health-related research concerning drowsiness detection. The study recommends Contrast-Limited Adaptive Histogram Equalization for near-infrared surveillance-based drowsiness detection systems. The activity demonstrates UDINUS's engagement in research addressing a computer vision challenge relevant to driver monitoring.