Multi-scale enhancement fusion for underwater sea cucumber images based on human visual system modelling

• Published in: Computers and electronics in agriculture 2020-08, Vol.175, p.105608, Article 105608
• Main Authors: Guo, Pengfei, Zeng, Delu, Tian, Yunbo, Liu, Shuangyin, Liu, Hantao, Li, Daoliang
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial intelligence; Human visual system modelling; Image contrast; Image enhancement; Image fusion; Image quality; Object recognition; Perception; Robotics; Sharpness; Underwater; Underwater image enhancement
• ISSN: 0168-1699; 1872-7107
• DOI: 10.1016/j.compag.2020.105608

•A multi-scale fusion enhancement method based on the human visual system modelling for underwater sea cucumber images is proposed.•The single image is enough for producing the enhanced image by the Laplacian image representation.•The human visual system related fusion gluing maps make all objects detection based tasks more clearly in our framework.•The human visual system modelling based algorithm is accurately accompanied with the underwater robot. The vision computing techniques are widely used in the marine industry. In order to improve the detection and recognition accuracy of artificial intelligence-robotics, we propose a novel underwater image enhancement algorithm, using a multi-scale fusion approach based on the properties of the human visual system. Our method fuses the results of underwater image enhancement algorithms that deal with the color-casting, sharpness and contrast degradation. The method is further weighted by a human visual system-based image structure map that combines Michaelson-like contrast map, saliency map, dark channel map and exposed map. The multi-scale fusion strategy is used to avoid the artifacts of sharpness blending based on Laplacian image representation. The results show that our algorithm can recover more detail information of dark regions and improve the overall visual quality of underwater images.