Low-light and hazy image enhancement using retinex theory and wavelet transform fusion

In contemporary times, image enhancement is playing an increasingly important role in image analysis and synthesis. This article introduces an algorithm that aims to enhance the visual information of low-light or hazy images while also improving quantitative metrics such as peak signal to noise ratio, discrete entropy, feature similarity, and structural index. The methodology leverages a multi-pronged approach involving illumination estimation, an enhanced coefficient for a filter to remove haze, along with the fusion of Discrete Wavelet Transform (DWT) tailored to fortify insufficiently illuminated images. In the initial phase, a computation of modified illumination is executed to rectify underexposed regions, thereby enhancing overall viewing quality. Following this, an inversion operation is applied to the image, setting the stage for an optimized de-hazing process aimed at eliminating haze artifacts. The subsequent integration of a DWT-based fusion strategy serves to distil salient features from both images, amalgamating them into a cohesive, enhanced output. Simulation results demonstrate a significant improvement over previously proposed algorithms in terms of visibility and quantitative performance.