Enhancing Driver Drowsiness Detection through GMM-Optimized CLAHE

This research investigates the integration of Gaussian Mixture Model (GMM) with Contrast Limited Adaptive Histogram Equalization (CLAHE) to enhance image contrast, specifically for driver drowsiness detection based on facial images. Traditional CLAHE employs a fixed clip limit for contrast enhancement, which may not adapt well to varying image characteristics. To address this, the proposed CLAHE+GMM method dynamically determines the optimal contrast limit using GMM clustering. This approach aims to achieve more adaptive and effective contrast enhancement. The evaluation metrics used in this study include Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Signal-to-Noise Ratio (SNR), and computation time, providing a comprehensive assessment of the methods' performance. Experimental results indicate that employing 2 clusters in CLAHE+GMM significantly improves MSE, PSNR, and SNR, demonstrating the method's effectiveness. Visual histogram comparisons and sample images further validate that CLAHE+GMM with 2 clusters closely aligns with the grayscale base image, ensuring stable and consistent enhancement. In conclusion, the CLAHE+GMM approach, particularly with 2 clusters, offers a promising solution for enhancing image contrast in driver drowsiness detection applications, contributing to improved safety in driving environments.