Denoising broadband CARS hyperspectral images on variance-stabilized data: a unified benchmark

Abstract

Significance Broadband coherent anti-Stokes Raman scattering (BCARS) enables label-free hyperspectral chemical imaging, but many signal components are small compared to the accompanying heteroscedastic noise, making optimal denoising crucial. Aim We seek to identify the highest-performing hyperspectral image denoising methods for BCARS. Approach We derive a physics-based, variance stabilization transform that yields nearly intensity-independent BCARS noise variance while preserving the dispersive Raman response. Within this framework, we compare six denoisers using a shared phase-retrieval pipeline: singular value decomposition, BM4D, Noise2Noise, deep image prior, self-supervised diffusion, and a cross-covariance-based singular value decomposition (CCV) introduced here. Evaluation uses five simulated cubes with exact ground truth and experimental measurements of glycerol, polystyrene beads, and Caenorhabditis elegans. Results In simulations, the variance stabilization transform reduces the ratio of maximum to minimum channel noise variance from approximately 72 to 1.2. Deep image prior and self-supervised diffusion denoisers achieve the highest mean foreground peak signal-to-noise ratios. CCV matches the mean foreground structural similarity of deep image prior with substantially shorter runtime and no per-cube rank tuning. Experimental comparisons reveal sample-dependent performance. Conclusions Physics-based variance stabilization provides a common framework for adapting and evaluating hyper-spectral denoisers for BCARS. CCV offers a practical balance of reconstruction fidelity, computational speed, and automated operation.