Haze is a phenomenon that is caused by the scattering of light due to atmospheric particles and aerosols, and it significantly degrades image quality by reducing contrast, color fidelity, and spatial resolution. This presents a critical challenge in applications such as aerial surveillance, autonomous navigation, and disaster management, where clear visuals are essential for effective monitoring and decision-making. Image dehazing, a computer vision technique, addresses this issue by restoring image clarity through the adjustment of imaging components. Conventional methods often fail due to limitations such as loss of detail, the presence of visual artifacts, and color distortion, which hinder their applicability in real-world scenarios. To overcome these challenges, this study integrates a robust deep neural network architecture designed to mitigate these drawbacks while delivering high-quality, visually appealing results with reduced computational overhead. The system incorporates boundary constraints and optimized convolutional layers to enhance the detection and removal of haze while preserving image details and color fidelity. The solution ensures visibility and reliability in critical fields reliant on high-quality image processing.