Physics-Based Haze Removal and Color Correction for Enhanced Underwater Image Visibility

Underwater imaging environments are inherently challenging due to complex optical propagation phenomena, primarily light scattering, wavelength-dependent absorption, and suspended particulate matter. These physical effects significantly attenuate light intensity and distort spectral characteristics, resulting in degraded image quality characterized by low contrast, color imbalance, reduced sharpness, and limited visibility range. Forward and backward scattering introduce veiling light and blur fine structural details, while differential absorption of longer wavelengths (particularly red light) leads to a dominant bluish-green color cast. Consequently, underwater images exhibit diminished color fidelity, poor edge definition, and suppressed texture information, which adversely impact object detection, segmentation, and visual interpretation tasks. To mitigate these degradations, this project proposes an advanced contrast and color enhancement framework for underwater haze removal based on physics-inspired and computational image processing techniques. The proposed methodology integrates dehazing models with adaptive color restoration strategies to compensate for wavelength attenuation and scattering effects. Specifically, the system incorporates the Dark Channel Prior (DCP)–based dehazing algorithm to estimate transmission maps and ambient light components, enabling effective removal of veiling haze. In addition, color correction mechanisms such as white balance adjustment, histogram equalization, and channel-wise compensation are employed to restore chromatic consistency and recover natural color representation. The processing pipeline begins with image normalization and noise suppression, followed by transmission estimation and scene radiance recovery. Contrast enhancement techniques, including adaptive histogram equalization (AHE) or contrast-limited adaptive histogram equalization (CLAHE), are applied to improve local contrast without amplifying noise. To further refine visual clarity, edge-preserving filtering methods such as guided filtering are utilized to smooth transmission maps while maintaining structural boundaries. The combined approach ensures enhanced visibility, improved depth perception, and restoration of object features that are otherwise obscured by turbidity. The developed system is implemented in MATLAB, leveraging its image processing toolbox for algorithm prototyping, matrix-based computation, and visualization. Performance evaluation is conducted on diverse underwater datasets captured under varying turbidity levels and lighting conditions. Quantitative assessment metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Underwater Image Quality Measure (UIQM), and contrast gain are used to validate enhancement performance. Experimental results demonstrate improved color balance, enhanced edge sharpness, and significant visibility restoration compared to conventional enhancement techniques. By improving visual interpretability and structural detail recovery, the proposed method supports applications in marine exploration, underwater robotics, coral reef monitoring, archaeological surveys, autonomous underwater vehicle (AUV) navigation, and environmental research. The framework provides a computationally efficient and adaptable solution suitable for both offline analysis and potential real-time underwater imaging systems.