Novel method for the characterization of bubble size distribution

Bubble size measurement using imaging technique is commonly used when visualization is possible. However, overlapped bubble images are commonly found. The absence of internal separating line in overlapped bubble images creates difficulty in automatic algorithm development for accurate bubble size measurement. Common methodologies such as manipulation of pixel connectivity, segmentation by morphological watersheds, and morphological approach such as granulometry, etc are unable to give accurate treatment to overlapped bubble images when there is high degree of overlapping. This work is aimed to propose a generalized method to determine the bubble size distribution of spherical, elliptical as well as irregular shape bubbles based on bubble images. A computer automated analysis is performed by converting overlapped bubbles into segmented arcs to represent individual bubbles. The segmented arcs are then used directly for size measurement without the need to reconstruct the contours of overlapped bubbles. With the elimination of the curve fitting step, no prior assumption of the bubble shape is needed and the computation time can be greatly reduced. A probabilistic model is also developed for the transformation of bubble size distribution measured using a fixed calibration scale into the actual bubble size distribution without the information of horizontal locations from the imaging device.