A Novel Multiscale Fusion Algorithm for Underwater Image Dehazing and Edge Preserving
Haoqian Huang, Yunfei Jin, Shian Sun, Mengdie Zhang · 2023
When we use visual devices for navigation and location in underwater environment, there is a great demand for image quality. However, underwater images have drawbacks such as severe color quality degradation, blurring, and low contrast, so it is necessary to preprocess the image to some extent. To address this problem, this paper proposes an underwater image dehazing algorithm based on a weight pyramid based multiscale image fusion framework. Specifically, for blue images, we use quadtree hierarchical search strategy and least squares filter to improve the acquisition process of background light in the dark channel prior method, making it more suitable for underwater images. For greenish images, we have improved saturation and value separately. Then the processed images are regarded as inputs to the multiscale pyramid to obtain the normalized weight mapping, making the image reconstructed. To verify the performance of proposed method, an experiment is conducted. As a result, the algorithm significantly corrects color degradation and the edges are also well preserved.