Since light is scattered and absorbed by water, underwater images have inherent degradation (e.g. hazing, color shift), consequently impeding the development of remotely operated vehicles. Towards this end, we propose a novel method, referred to as Best o f Bo th Worlds (Boths). With parameters of only 0.0064M, Boths can be considered a super lightweight neural network for underwater image enhancement. On the whole, it has three levels: Structure and Detail Features; Pixel and Channel Dimensions; High and Low Frequency Information. Each of these three levels represents ”Best of Both Worlds”. Initially, by interacting with structure and detail features, Boths can focus on these two aspects at the same time. Further, our network can simultaneously consider channel and pixel dimensions through 3D attention learning, which is more similar to human visual perception. Lastly, the proposed model can focus on high and low frequency information, through a novel loss function based on the wavelet transforms. Upon subsequent analysis and evaluation, Boths has shown superior performance compared with state-of-the-art methods. Our models and datasets are publicly available at: https://github.com/perseveranceLX/Boths.

Underwater images often exhibit strong color distortion and hazing due to various degradation factors. A majority of algorithms provide color correction for underwater images, but the color is not as vivid as it could be. In order to solve this problem, we propose a method known as adaptive underwater image enhancement guided by generalized imaging components (AUIE-GIC). To the best of our knowledge, this is the first method of utilizing deep learning to develop a generalized imaging model. The proposed method contains two stages: component generation and component guided learning. In the first stage, we obtain the components of a learning-based generalized imaging model. In the second stage, attenuation and transmission features are used to adjust color and semantic information. We propose the attenuation attention module (AAM), and transmission attention module (TAM), which can highlight heavily degraded areas. Finally, the refined network is able to restore underwater images that are rich in semantics and vivid in color. Based on further evaluation and analysis, AUIE-GIC is demonstrated to provide superior performance when compared with state-of-the-art (SOTA) methods. We make the source code available at https://github.com/perseveranceLX/AUIE-GIC .

Underwater images often suffer from severe color distortion due to the challenging imaging environment. Underwater image enhancement (UIE) techniques have been developed to recover clear images, laying the foundation for various underwater researches. However, existing UIE methods tend to produce fixed results without considering individual preferences for different color tones. And there is no dataset with ground truth in different tones. Therefore, we came up with the possibility of using the currently popular multi-modal methods to control the color tone of enhanced images. This paper proposes a method for generating underwater enhanced images with cold, warm, and normal tones using multi-modal information supervision (MM-UIE). Firstly, we leverage the relationship between text prompt and images to supervise the generation of cold or warm images. Additionally, we introduce a 6D color operator, which not only enhances the tone control of underwater images but also serves as a bridge between different tone images. Finally, we also found that multimodal supervision methods can not only control the color tone of underwater images, but also improve the quality of underwater image generation. Experimental results demonstrate the superior performance of our method compared to state-of-the-art (SOTA) techniques. Our codes will be publicly available at: https://github.com/perseveranceLX/MM-UIE.

Stereo images are widely used in many scenarios such as autonomous driving and stereoscopic display. Stereo image super-resolution is a common technique to obtain high-definition stereo images. However, existing methods are mainly designed for fixed scales and cannot smoothly support arbitrary upscaling. To address these problems, this paper proposes a Stereo ImaGe NeurAl Learner (SIGNAL) based on implicit neural representation, which consists of three parts. First, a feature extraction module leveraging the iPASSR architecture to robustly extract hierarchical features from low-resolution stereo images. Second, a novel stereo implicit representation module that explicitly models the inherent geometric consistency and contextual relationships between the two views, enhancing the latent features through view-specific and cross-view interactions. Third, a cross-view decoding module designed to fully exploit the complementary information across views during the image reconstruction process. Extensive experiments across diverse datasets demonstrate that the proposed SIGNAL framework achieves superior performance. The results underscore the effectiveness of our approach in handling arbitrary scaling factors while preserving visual fidelity in stereo image restoration.

In the challenging underwater imaging environment, underwater images often suffer from mixed degradations. Current methods are often insufficient to comprehensively address the issues of light, color, and contrast. To tackle this, we propose a novel text-image joint learning approach called multi-color attribute information fusion (MCAIF), which consists of four enhancement branches for brightness, color, contrast, and fusion, respectively. Initially, we use paired text prompts to supervise the learning of branches in a contrastive manner. To fully use the interdependence of four branches, an implicit interaction module is proposed to facilitate color attribute information collaborative communication. In view of the significant color degradation in underwater images, neural operators are presented to enhance the input image through deep filtering, which includes not only the raw channels, but also the assisted channels by utilizing a learning-based white balance model. Extensive experiments are conducted on various datasets, including the T200 test set that we collected to evaluate the overall performance of UIE algorithms. Experimental results across datasets demonstrate that our approach outperforms the state-of-the-art methods, producing visually pleasing results that align markedly better with human perception.