Optimized self-registration method for pseudo-depolarization calibration in Mueller matrix imaging

Mueller matrix polarimetry can effectively extract structural characteristics of tissue specimens by analyzing how tissue structures modulate the polarization state of incident light. As a key parameter derived from Mueller matrix for quantifying tissue properties, depolarization exhibits high sensitivity to diverse microstructural features. However, in dual-rotating retarder Mueller matrix imaging systems, image artifacts induced by beam drift or rotation errors can result in pseudo-depolarization effects, severely interfering with accurate depolarization parameter calculations. In this study, we propose an optimized self-registration method to mitigate pseudo-depolarization effects. We employ a low-pass filter to suppress high-frequency noise in constant-step rotating Mueller matrix polarimetry imaging. Registration displacement vectors are then optimized via backpropagation and applied to raw images to correct image artifacts. Experiments on pumpkin stem tissue demonstrate the feasibility of our method, which effectively enhances the accuracy of depolarization parameter calculations. Texture analysis further confirms an additional advantage of the optimized self-registration method to enhance texture structure information in certain Mueller matrix parameters, enabling more precise characterization of biological tissue microstructures.