Dynamic speckle imaging with SVD compression

Abstract

Abstract Dynamic speckle imaging (DSI) of areas with different speed of processes ongoing in industrial or biological objects relies on statistical processing of a large number of images of the speckle patterns formed on the objects surface under laser illumination. The DSI visualizes the speed spatial distribution as an activity map. We propose compression of the raw DSI data by applying singular value decomposition (SVD). A specific feature of speckle images for DSI is lack of a structure with areas of close intensity values. The gain from the direct SVD application may be modest in cases when a great number of non-zero singular values is needed to build an activity map comparable in quality to the ground truth map from bitmap images. For higher compression, we propose SVD to be applied to the 2D arrays containing the differences between the successive images. The SVD compression has been verified by using synthetic and experimental data.