Efficient wavelet-based optical image dehazing technique

Outdoor scene images captured in bad weather conditions frequently have limited visibility and little contrast, with these issues varying across the images. Adverse weather can hinder the clarity of outdoor images, resulting in a decline in their visual quality. These images can face multiple challenges, including haze caused by atmospheric particles like dust and smoke. Videos captured under such conditions are also affected by noticeable quality degradation, which includes reduced contrast and loss of details. Image dehazing is the name of the image processing task intended to lessen this impact. This paper introduces an efficient wavelet-based optical image dehazing technique that eliminates visual deterioration caused by haze without depending on the inversion of the physical model of haze formation, while still adhering to its primary underlying assumptions. Consequently, the proposed wavelet-based optical image dehazing technique eliminates the necessity for depth estimation in the scene and the associated expensive depth map refinement processes. An efficient multi-scale correlation wavelet approach in the frequency domain is adopted to resolve the problem of image dehazing and denoising in this paper. First, we simultaneously utilize soft-thresholding and median denoising operations to suppress noise for adaptively boosting the texture details in the high-frequency regions. After that, the wavelet reconstruction of the enhanced high-frequency parts and the recovered low-frequency part is used to properly reconstruct the denoised image. Finally, the denoised image is artificially underexposed through a series of gamma correction operations. A collection of multiple exposed images is fused into a haze-free result using a multi-scale Laplacian blending approach. In-depth qualitative and quantitative experimental analysis is provided. The experimental results show that the fusion of artificially under-exposed images works very well for dehazing in scenes with challenging conditions, while other current methods fail to provide high-quality dehazing effects. Furthermore, higher perceptual visibility is not the only target of the proposed technique, but more texture details are kept and less noise is achieved. Extensive experiments have demonstrated that our technique is comparable and even superior to the classical competing techniques. On average, the achieved entropy of dehazed images is 7.507, the fog aware density evaluator (FADE) is 0.289, and the visual contrast measure (VCM) is 93.289.