Data compression and improved registration for laser speckle contrast imaging of rodent brains.

Single-frame blood flow maps from laser speckle contrast imaging (LSCI) contain high spatiotemporal variation that obscures high spatial-frequency vascular features, making precise image registration for signal amplification challenging. In this work, novel bivariate standardized moment filters (BSMFs) were used to provide stable measures of vessel edge location, permitting more robust LSCI registration. Relatedly, BSMFs enabled the stable reconstruction of vessel edges from sparsely distributed blood flow map outliers, which were found to retain most of the temporal dynamics. Consequently, data discarding and BSMF-based reconstruction enable efficient real-time quantitative LSCI data compression. Smaller LSCI-kernels produced log-normal blood flow distributions, enhancing sparse-to-dense inference.