Fig. 1. Parameter estimation algorithm based on empirical mode decomposition (EMD). (a) Spectral auto-correlation; (b) frequency distribution of ideal parameter estimation; (c) frequency distribution of actual parameter estimation and its EMD result

Fig. 2. Flow chart of EMD parameter estimation

Fig. 3. Calculation flow chart of dispersion index

Fig. 4. Dispersion index. (a) Schematic of interval partition: (b) schematic of weight distribution function and value range of function; (c) dispersion index of EMD waveform of Fig.4 (d) and Fig.4 (e); (d) initial phase (from left to right and from top to bottom) estimated by averaging original images at the first 1, 3, 5, 7, 9, and 11 time points after exciting the same sample for many times; (e) modulation depth (from left to right and from top to bottom) estimated by averaging original images at the first 1, 3, 5, 7, 9, and 11 time points after exciting the same sample for many times

Fig. 5. Flow chart of VST-VBM3D denoising algorithm

Fig. 6. Comparison of parameter estimation before and after denoising. (a) Cytoskeletal actin microfilaments; (b) comparison of EMD waveform estimated with initial phase before and after denoising in three directions; (c) comparison of EMD waveform estimated with modulation depth before and after denoising in three directions

Fig. 7. Comparison of reconstruction results before and after denoising. (a) Ground truth image; (b) reconstruction result without denoising; (c) reconstruction result after denoising; (d) locally magnified images of Figs. 7(a), (b), and (c)