Preprocessing of raw data for quality enhancement of the pointwise dynamic speckle analysis

The pointwise intensity-based processing of time-correlated speckle patterns allows for detection of activity in three-dimensional objects by building a two-dimensional map of a certain statistical parameter. We have developed in this work reliable approaches for high-quality visualization of the activity map through normalization and preprocessing of the captured raw data. As a first task, we analyzed statistical behavior of correlation-based estimates to show erroneous determination of activity under non-uniform illumination or varying reflectivity across the object surface for non-normalized algorithms and wrong detection of zero-activity regions for the normalized algorithms. Next, we proposed solution for the non-uniform illumination issue by using the sum of the speckle patterns at a given time lag for normalization and by introducing a flexible threshold to form binary patterns. We checked the option of the spatial smoothing of the raw data for activity map visualization enhancement and proved that smoothing in the temporal domain was more effective. Efficiency of the proposed preprocessing for data captured at uniform and non-uniform illumination was demonstrated on synthetic and experimental speckle patterns.