Image Threshold Segmentation Without Criterion Function-Optimal Nectar Source Algorithm

At present, the objects of the medical image segmentation are mainly about various cells,tissues and the organs. Human cell image processing which is the most important anddifficult part，especially it has an important influence on segmenting cell area accurately,disease diagnosis, cell quantitative analysis and so on. The most important image processingstep is the image segmentation, since the results will affect the quality of the follow–up work.Hence the study of image segmentation methods is very important.This paper introduces some types of the image segmentation methods and focuses on thethreshold segmentation. It gets the conclusion through the comparison and analysis: now,most of the threshold segmentation methods get a criterion according to the relatedinformation or the different theory. The criterion was regarded as the target function ofoptimization problem, so the image threshold segmentation was transformed to anoptimization procedure. Unlike other image threshold algorithms, it try to find an imagethreshold algorithm without a criterion. According to the flower-counting principle of honeybees in Artificial Bee Colony algorithm, it assume that there exists an optimal nectar sourceand the flower-counting of the honey bees is consistent and stable. After defining the modelof flower counting and threshold updating, the coding and colony initialization, thecalculation of criterion, the selection mechanism and the realization of termination conditions,it proposed an image threshold algorithm without a criterion, namely Optimal Nectar SourceAlgorithm (ONSA). Experimental results show that the optimal nectar source does exist, thepopulation is strongly attracted by the nectar source. The nectar source is the most stablefactor during the flower-counting. In the algorithm feasibility experiment, it compared theother algorithms with the ONSA on the Lenna picture, from this we know that the ONSA isfeasible to be used in the image segmentation.On this basis, it promotes the algorithm to the medical cell images from using thealgorithm to segment the cell images, and the test picture usually be used. The related testresults show that the proposed algorithm is valuable.