Non-invasive real-time imaging through turbid water based on coherent detection

This dataset is strictly associated with the paper titled《Non invasive real time imaging through turbid water based on coherent detection》. It exclusively contains key images used in the paper to validate core technologies and present experimental results, excluding raw experimental data or additional derived content. All images directly correspond to the figures in the paper (e.g., Fig. 1 to Fig. 14 and appendix figures), fully reproducing the research logic and conclusions of the study. The dataset content is categorized according to the experimental dimensions of the paper, including comparative images of different gating methods (traditional time gating [TG], coherent gating [CG], and coherent time gating [CTG]), holograms from the CTG method and their processed results via the Digital Holographic Reconstruction Algorithm (DHRA), comparative images of the physics-informed two-step neural network (SHARPNet) before and after optimization, verification images for spatial and depth resolution, and supporting images for experimental rationality (e.g., spatial frequency [SF] analysis graphs, pixel intensity histograms, and interference fringe images for coherence length measurement). Each image implicitly includes annotations of core experimental parameters from the paper, such as a laser wavelength of 532 nm, a pulse width of 1.1 ns, a scattering medium of fat emulsion suspension (with an attenuation length [AL] ranging from 6 to 12), and a gated ICCD camera (Andor iStar 334T) with a gate width of 2 ns, which are fully consistent with the experimental design of the paper.The core value of this dataset lies in its intuitive presentation of the key research findings of the paper. These include the advantages of the CTG method over TG and CG—achieving an imaging distance of up to 12 AL (2–3 AL longer than TG) and an 8–9 dB improvement in Peak Signal-to-Noise Ratio (PSNR)—as well as the further enhancement of PSNR to 12–13 dB after optimization by SHARPNet, and the CTG method’s 1 mm spatial resolution and millimeter-level depth resolution (with an error of approximately 0.2 mm). All images are original collected and processed results from the paper’s experiments without any modifications, making them directly applicable for academic communication of the paper’s achievements, cross-comparative research, and teaching demonstrations. For instance, the comparative images of different gating methods clearly show that under 12 AL turbid conditions, TG fails to resolve target contours, CG results in blurred details, while CTG can clearly present target details. The SHARPNet optimization images annotate the differences in image quality under three modes (no pre-training, pre-training only, and pre-training + fine-tuning), verifying the effectiveness of the two-step training (pre-training on a simulated dataset + fine-tuning with approximately 200 real data pairs), which is highly aligned with the paper’s research goal of "reducing reliance on large-scale real-world data".When citing this dataset, the full title of the corresponding paper《Non invasive real time imaging through turbid water based on coherent detection》and the specific figure number of the image in the paper (e.g., "Fig. 8", "Fig. 11", "Appendix Fig. 12") must be clearly indicated. For access to detailed experimental parameter derivations, algorithm codes, or the full text of the paper corresponding to the images, please contact the paper’s authors (E-mails: [e-mail]; [e-mail]). All content of this dataset strictly adheres to the research scope of the paper and does not involve any extended information beyond the paper, ensuring consistency and accuracy with the paper’s achievements. It provides a paper-experiment-based visual reference sample for fields such as underwater optical imaging, imaging through scattering media, and deep learning-based image enhancement.