Image threshold segmentation in the colonies based on difference method of statistical curve
In many fields such as food, pharmacy, medicine, quality and technical supervision and so on, we usually need colony counting. Today it is often realized by using computer image processing technology. This method can decrease the huge labor intensity of the traditional operator in artificial counting, and does not need too much prior knowledge of the colony recognition, Yang Fan (2011). With people using image processing techniques to count the images of the colonies, avoiding interference background noise on image processing as far as possible and determining the threshold value have become a big problem because of the colony background noise making statistical accuracy very dierent. So far, the researchers made a lot of researches on how to get the threshold of the image, and put forward a lot of classic methods, such as maximum between-cluster variance method, Otsu, N (1979) (also called Da-Jin method, OTSU method), maximum entropy value method, Kapur, J.N & Sahoo, P.K (1985), maximum correlation method Yen, J.C & Chang, F.J (1995), 2-mode method, Lee, S & Chung, S.A (1990), p parameters method, Doyle, W (1962) and so on. Other scholars also put forward many corresponding algorithms in dierent image processing fields. However, the interference of background noise remains serious with these methods. In the field of the cell counting, Dahle, J (2004) and his colleagues used the OTSU method to get the threshold value in his paper named for Automated counting
of mammalian cell colonies by means of a flatbed scanner and image processing. Correspondingly, the OTSU method which we can see in the papers” An Automated Bacterial Colony Counting System”, Zhang, C.C & Chen, W.B (2008) and” An automated bacterial colony counting and classification system”, Chen, W.B & Zhang, C (2009) is also used to get that value in the field of the colony counting. According to the current literatures, researchers have not the corresponding studies on the image threshold segmentation algorithm of the colonies. Therefore, this paper proposes a dierent method of statistical curve after a study in which we discussed the relation between the colony counting result and corresponding Threshold.