Fig. 8![\"Fig.](\"https:\/\/www.newsbeep.com\/nz\/wp-content\/uploads\/2026\/08\/41598_2026_65714_Fig8_HTML.png\")
<\/p>\n
The scanned copy of the formal ethical approval report for the endoscopic video dataset.<\/p>\n
Appendix B: Additional experimental details<\/p>\n
This appendix provides additional details on the clinical dataset, data splitting strategy, synthetic degradation settings, training configurations, downstream VSR training settings, and t-SNE visualization settings used in the experiments.<\/p>\n
Synthetic dataset information<\/p>\n
The information of the dataset used in \u201cEvaluating simulated LR images\u201d and \u201cVSR on the synthetic dataset\u201d sections is shown in Table\u00a09<\/a>.<\/p>\n Table 9 Statistics of the collected clinical endoscopic video dataset.Data splitting strategy<\/p>\n The data splitting strategy used in the experiments is summarized in Table\u00a010<\/a>. We divided the 200 selected HR video sequences into two non-overlapping subsets. The first subset was used as the source-domain HR data for degradation learning. The second subset was processed by different synthetic degradation settings to construct the target-domain LR data for adversarial degradation learning. In addition, 20 extra video sequences were reserved for testing. The split was performed at the video-sequence level to avoid information leakage.<\/p>\n Table 10 Data splitting strategy used in the experiments.Synthetic degradation settings<\/p>\n The detailed synthetic degradation settings are reported in Table\u00a011<\/a>. This synthetic benchmark is designed to cover several typical degradation factors that may appear in endoscopic videos, including interpolation degradation, isotropic Gaussian blur, direction-dependent anisotropic blur, motion blur, and noise. All degradation settings are conducted under the \\(\\times 4\\) upscaling factor.<\/p>\n Table 11 Synthetic degradation settings used in the experiments.UDLM training settings<\/p>\n The training settings of the proposed UDLM are summarized in Table\u00a012<\/a>. During training, each input sample consists of a video clip with 10 consecutive frames. We crop HR patches of size \\(256 \\times 256\\) from the original HR frames. Under the \\(\\times 4\\) super-resolution setting, the corresponding LR patch size is \\(64 \\times 64\\). The model is first trained with the low-frequency loss during the warm-up stage. After that, the adaptive data loss is introduced, and the adaptive kernel is periodically updated according to the current behavior of the degradation generator.<\/p>\n Table 12 Training settings of the proposed UDLM.Loss weights<\/p>\n The loss weights used for UDLM training are listed in Table\u00a013<\/a>. The low-frequency loss and adaptive data loss share the same data-loss weight. After the warm-up stage, the data constraint is switched from the low-frequency loss to the adaptive data loss.<\/p>\n Table 13 Loss weights used for UDLM training.Training settings of downstream VSR models<\/p>\n The training settings of the downstream VSR models are summarized in Table\u00a014<\/a>. For fair comparison, all pseudo-paired data generated by different degradation learning methods are used to train the downstream VSR models under the same configuration.<\/p>\n Table 14 Training settings of downstream VSR models.Real clinical dataset statistics<\/p>\n The statistics of the real clinical endoscopic dataset used in \u201cPerformance evaluation in real-world endoscopic scenarios\u201d section are shown in Table\u00a015<\/a>.<\/p>\n Table 15 Statistics of the real clinical endoscopic dataset used in \u201cPerformance evaluation in real-world endoscopic scenarios\u201d section.t-SNE visualization settings<\/p>\n