Thresholding which is a crucial part of image segmentation has been known to help in the unification of objects and moreover, it will attenuate the noise within the regions of interest which ultimately, leads to significantly better results [1, 2]. Rubber latex concentrate, which is essentially an intermediate state between suspensions and solutions, is the raw material for making many common household and industrial products. A common measure of its quality is the mechanical stability time, which is defined as the time at the first onset of flocculation when the latex is subjected to the physical stress from rotating it at speeds of 14,000 rev/min. Nevertheless, the common approach to determine this, based on the widely adopted ISO 35 & ASTM 1076 standards[3, 4] relies heavily on the experience, eye sight and analytical skills of the human operator when presented with the extracted samples taken at regular time intervals[5]. In this paper, we propose a new approach to determine the mechanical stability time of the sample but as it involves particle enumeration, we will show how multi-thresholding based on Artificial Bee Colony (ABC) can significantly improve the accuracy of this enumeration.