Multicluster Kernel Intuitionistic Fuzzy C-Means and State Transition Algorithm: Framework for Low Light Segmentation Imaging of Tuberculosis Bacilli Base of Semisupervised Approach

Annotated images are required as ground truth for deep learning. Bacterial segmentation is requires a lot of time and effort when done manually. The autosegmentation task gets more challenging since bacterial images contain low light properties, which have an important effect on auto annotating tasks. In order to solve this issue, we present a system that includes a fuzzy-based clustering method that enhances bacterial object segmentation performance by utilizing the multicluster idea. The State Transform Algorithm (STA) is used to obtain starting centroids in order to increase stability, because the Kernel of Intuitionistic Fuzzy C- Means (KIFCM) is sensitive to starting centroids and hence sensitive to being stuck in local optima. The accuracy of KIFCM-STA with bicluster is poor in low-light images. To boost performance, the multicluster technique (MKIFCM-STA) is presented as a continuation hybrid of KIFCM-STA. This framework allows for the ideal amount of clusters (Silhouette) and cluster ranking to provide clusters containing bacterial objects (Topsis). In order to compare our method against four prior approaches (IFCM, KFCM, KIFCM, and KIFCM-STA), we compare its qualitative aspects (visualization of images) and quantitative aspects (average IOU, Dice, HD, ASD, and Accuracy). In low light image clustering tasks, our model significantly improves and achieves great results in terms of accuracy, with a score of 89.438%. This accomplishment highlights how crucial it is that our framework tackles the problem of low-light image clustering in images of bacteria, eventually improving the image auto-annotation procedure.