Zero-order suppression in off-axis digital holography combining optimized non-local means filtering and a variational image decomposition model

In off-axis digital holography, phase reconstruction typically involves extracting the +1-order term spectrum, which carries the phase information of the sample. However, interference caused by the zero-order term can reduce the accuracy of this extraction. This study proposes a novel variational image decomposition model, to our knowledge, that effectively suppresses the zero-order term. An integral image is first constructed to store the cumulative sum of squared pixel differences to improve the computational efficiency of the non-local means (NLM) filter in removing random noise. The stored values are used to rapidly compute the similarity of neighboring windows by avoiding repeated summation operations, which reduces the computation time of NLM filtering. Subsequently, the optimized NLM filter is employed to remove random noise from the hologram. The denoised hologram is decomposed into structural and textural components using the TV-G algorithm. The structural component corresponds to the zero-order term while the textural component retains the interference fringes. Based on the decomposition results, the zero-order term is effectively suppressed, and the textural component is used for phase reconstruction. Additionally, the dynamic fluctuation phase maps of water droplet evaporation were retrieved to demonstrate the real-time capability of the method.