Previous studies on image processing techniques for particle sizing have focused mostly on a single particle or weakly overlapped particles. In the present work, the image processing algorithm for particle size measurement has been improved to process heavily overlapped spherical particle images. The algorithm consists of three major steps; detection of boundaries of the particle clusters, identification of the individual particles, and false circle elimination. For the first step, the Sobel operator (using gray-level gradient) and the boundary thinning process were adopted, and compared with the gray-level thresholding method that has been widely adopted. In the second step, the Hough transform technique was used. The Hough transform is an algorithm to detect parametric curves such as straight lines or circles which can be represented by several parameters. Then, to improve the measurement reliability, the process of eliminating the false circles from the "particle-like" ones was added. The improved algorithm was tested by processing an image frame which contains heavily overlapped spherical particles. The results showed that both the performances of detecting the overlapped images and separating the element particles from them were satisfactory.