Structural Prior-Guided Weighted Low-Rank Denoising for Short-Wave Infrared Star Images

In ground-based short-wave infrared (SWIR) astronomical observations, temperature drift in the detector readout circuit often introduces nonlinear, spatially non-uniform stripe noise together with Gaussian noise, making weak stellar targets easily submerged and difficult to detect. To address this challenge, we propose a structurally guided weighted low-rank denoising method for infrared star images. Going beyond traditional spatial filtering and standard low-rank decomposition, the proposed method integrates physical priors with mathematical optimization into a unified framework. First, the point spread function (PSF) characteristics of stellar targets are used to construct a hierarchical structural filter, which is further transformed into adaptive prior weights. This design preserves weak-target energy while suppressing noise during iterative optimization. Second, by exploiting the global spatial correlation of the image, residual stripes and the background are jointly modeled as a low-rank component for effective separation. Finally, Bilateral Random Projection (BRP) is introduced to accelerate the weighted soft-thresholding iterations. Experiments on real ground-based observation data, together with ablation studies and sensitivity analyses, demonstrate that the proposed method effectively suppresses structured stripe interference while preserving weak stellar targets under low-SNR conditions. In addition, the acceleration module further improves computational efficiency, making the framework more suitable for practical real-time processing.