The field of image processing is moving towards more sophisticated and specialized techniques, with a focus on addressing specific challenges such as low-light conditions, ultra-wide-field imaging, and spectra processing. Recent developments have introduced innovative architectures and methods, including the integration of illumination and semantic priors, frequency-aware self-supervised learning, and mutual information-based networks. These advancements have shown promising results in enhancing image quality, preserving pathological details, and improving disease diagnosis performance. Noteworthy papers include: ISALux, which proposes a novel transformer-based approach for low-light image enhancement, and A Frequency-Aware Self-Supervised Learning for Ultra-Wide-Field Image Enhancement, which introduces a frequency-decoupled image deblurring module. Additionally, Learning Binary Sampling Patterns for Single-Pixel Imaging using Bilevel Optimisation demonstrates superior reconstruction performance compared to baseline methods, and Sky Background Building of Multi-objective Fiber spectra Based on Mutual Information Network proposes a sky background estimation model that utilizes spectra from all fibers in the plate.