The ability of microorganisms to colonize different types of surfaces and interact chemically and biologically with the medium they colonize has been object of study for years. An example of this is the large number of projects and publications on the effect of colonizations on human health, on the bioprotection of rocky surfaces in structures of cultural interest, on biopesticides, on the degradation of plastics in marine ecosystems and on biofertilization and biostimulation of plants. Techniques such as FISH (Fluorescent in situ hybridization) or CARD-FISH (catalyzed-reporter deposition- FISH), and the subsequent fluorescence or confocal microscopy, make up a synergy of available methods to detect and monitor the presence of microorganisms of interest in different environments. , as is the case with the roots of plants. This study uses a collection of images obtained from the doctoral project Promotion of plant growth of Bacillus subtilis EA-CB0575, rhizospheric colonization and genomic and biochemical potential, in which the colonization of the strain Bacillus subtilis EA-CB0575, a promoter of plant growth in crops of economic interest. The purpose of this collection of images was to evaluate the presence of the stated microorganism in the roots of the evaluated plant species (banana, Musa AAA var. Williams and tomato, Lycospersicum esculentum var. Chonto) and to monitor Bacillus subtilis through the use of a or several fluorescence probes; generating images where the cells of interest, fluorescing due to the process carried out, are clearly contrasted with the black or low fluorescence background. This investigation was divided into two phases; the first consisted of the design of an algorithm, executed by the work team. This resulting algorithm was called MSA, and its purpose is to segment images with the aforementioned fluorescence characteristics. The second phase consisted of the segmentation of the images by the supervised algorithms RATS, RATS L, Edge Detection, LOCAL, Isodata, and the unsupervised training for the definition of parameters of the Canny algorithm, some of them implemented in software already existing and available for microscopy image analysis as DAIME and Image J. The algorithm that obtained the smallest difference between the cell count by the expert and the segmentation of the regions of interest was the Canny edge detection algorithm with an RMSE value of 67.8, on the other hand, under the error metric MAE and the measure of accuracy the algorithm that uses a global threshold value RATS obtains the best performance, the highest precision is obtained by the Canny algorithm (84%), and the highest sensitivity (43%) is obtained by the Detection algorithm. edges. The performance of the MSA algorithm was positive, since in all the defined measures it was above the average, the best performance was obtained in the precision measure (71.4%). During the development of this research, the subjectivity and bias that exists in the analysis of the images was evidenced, since for the same expert it is difficult to replicate the results in the cell count made previously. For this reason, the methodology proposed here is an alternative to eliminate subjectivity and turn the analysis of microscopy images into a process that allows reproducibility of the results.