Refinement of particle detection by the hybrid method in digital in-line holography

Abstract

Digital in-line holography provides simultaneous particle size and three-dimensional position measurements. In general, the measurement accuracy varies locally, and tends to decrease where particles are closely spaced, due to noise resulting from diffraction by adjacent particles. Aggravating the situation is the identification of transversely adjoining particles as a single particle, which introduces significant errors in both size and position measurements. Here, we develop a refinement procedure that distinguishes such erroneous particles from accurately detected ones and further separates individual particles. Effectiveness of the refinement is characterized using simulations, experimental holograms of calibration fields, and a few practical applications to liquid breakup. Significant improvements in the accuracy of the measured particle sizes, positions, and displacements confirm the usefulness of the proposed method.

Automated droplet size distribution measurements using digital inline holography

“…Next, the hologram was reconstructed by convolution with the Rayleigh-Sommerfeld diffraction kernel to obtain the 3D intensity field containing all of the droplets (Figure 3c shows a minimum intensity projection of this field). Subsequently, we implemented the hybrid algorithm described by Gao et al (2014) to automatically segment the minimum intensity projection to identify the droplets, and further refined their position in the longitudinal direction (along the laser beam) using a sharpness based focus metric (Gao et al, 2014). Comparing the result of the refinement (Figure 3c) with the projected image, we can clearly observe the sharp edges that help accurately measure the size and shape of the each droplet.…”

A hybrid image processing method for measuring 3D bubble distribution using digital inline holography

“…10, the edge detection results of the segmented bubbles are superimposed onto the depth map of individual pixels for measuring the longitudinal location of individual large bubbles/clusters. Comparing to previous method that determines the degree of focus through intensity gradient (Guildenbecher et al 2013, andGao et al 2014), the focus metric approach employed in the current study eliminates the need for manually choosing window of specific size to calculate intensity gradient using Sobel operator. In comparison with other wavelet-based focus metric approaches (e.g., Wu et al 2014), the Haar wavelet adopted in our study uses single-pixel support window which enables more accurate characterization of the intensity gradient across the bubble edge, particularly for small bubbles.…”

Machine learning holography for 3D particle field imaging

“…The depth of particles can also be determined through quantification of the level of particle focus using various methods (e.g. Gao et al 2014, Wu et al 2014. For example, Gao et al (2014) used the pixel intensity gradient to quantify sharpness of the particle edge (which determines the particle depth), and measured particle size using the minimum intensity metric.…”

“…Gao et al 2014, Wu et al 2014. For example, Gao et al (2014) used the pixel intensity gradient to quantify sharpness of the particle edge (which determines the particle depth), and measured particle size using the minimum intensity metric. This method has been employed in various applications such as measurements of spherical solid particles in quiescent flow (Gao et al 2014) and spray measurements in a wind tunnel (Kumar et al 2019).…”