A novel environment-adaptive dual-light image enhancement framework for marine oil spill detection

Ocean oil spills pose a severe threat to the marine environment. This research addresses the significant challenge of detecting low-contrast oil spills on the sea surface, a problem exacerbated by the presence of specular reflections from sunlight in visible light images and thermal noise in infrared images. A novel environment-adaptive dual-light image enhancement framework is proposed for marine oil spill detection. Firstly, an improved Criminisi sun glint inpainting algorithm is proposed to eliminate the effects of sun glint regions in visible light images. As the oil spill regions in infrared images can be distorted by thermal noise and background interference, a novel Difference of Gaussian Weighted Guided Image Filtering (DoGWGIF) enhancement algorithm is then created to enhance local detail clarity and significantly increase the overall contrast of the oil spill targets, thereby improving the oil spill detection ability in the infrared images. The proposed algorithms were validated through experiments conducted on marine sun glint regions and low-contrast infrared areas, demonstrating their effectiveness. The performance index MIoU went up 1.64 % and 0.54 % for visible light images and infrared images, respectively. •An enhanced version of the Criminisi inpainting algorithm is proposed.•A new infrared image enhancement method is developed.