Automatic measurement of surface tension from noisy images using a component labeling method

Automatic surface tension measurement of turbid liquids using a bubble method (e.g., pendant/captive bubble) or interfacial tension measurement of some liquid–liquid systems necessitates an image analysis scheme robust against noise. Measuring the surface tension of lung surfactant–polymer systems using the combination of axisymmetric drop shape analysis (ADSA) and a captive bubble is such an example. We have recently developed a sophisticated image analysis scheme featuring the combination of the Canny edge detector and a novel edge smoothing technique called axisymmetric liquid fluid interfaces-smoothing (ALFI-S). The Canny edge detector is highly noise-resistant and ALFI-S is efficient in further removing noise close to the bubble profile (i.e., adhering noise). However, the procedure of eliminating noise far away from the bubble profile (i.e., isolated noise) is still not satisfactory in the previous scheme. Here a component labeling method is developed to automatically remove isolated noise. Component labeling refers to the process of detecting connected objects in a digital image. Once labeling is completed, information, such as area, location and pixel intensity, of these separate objects can be obtained. Hence, component labeling can be used as a region-based segmentation technique that is capable of differentiating the bubble and the isolated noise. The new component labeling-based noise reduction method is used to analyze captive bubble images with different amounts of noise. It is found that with the help of the labeling procedure smooth edges can be detected with a simple set of user-specified parameters even from images with extensive noise. Combined with the previous image analysis scheme, the component labeling method significantly improves the effectiveness and reliability of ADSA in automatically measuring surface tension from noisy images.