In this study, an image processing program was developed to deduce parameters of the elliptic shapes of liquid droplets using the randomized Hough transform and parameter decomposition. Partially overlapping droplets were separated during a shape detection process. Five parameters of an ellipse were calculated by dividing these parameters into two groups, namely, the center position and the remaining parameters (major and minor axis lengths and axis orientation), which is known as the parameter decomposition method. Candidate ellipse centers were determined by using the geometric properties of ellipses. The standard and randomized Hough transforms were applied to locate centers and to determine the values of the remaining parameters in two and three dimensions, respectively, instead of the five dimensions originally required. Best ellipse estimations were then determined for several candidate ellipses by comparing original droplets with ellipses constructed using the estimated parameters. The developed program was applied to simulated overlapping ellipses, to real overlapping droplets, and finally, to real spray droplets. The results obtained using the developed method were satisfactory, unless there were inherent problems with the original images. This method can be also used as an effective means of separating overlapping small particles.