Malaria is a serious disease that especially affects developing countries. Resource constraints in developing countries make automated and expert replacement systems particularly important. The machine learning algorithms were studied to provide this automation where malaria parasites are detected and counted. The machine learning pipeline for malaria parasitemia classification includes segmentation and classification steps for end-to-end assessment. In the segmentation step, red blood cells (RBCs) are segmented for individual evaluation of RBCs. However, it is a challenging task in case of overlapping RBCs. In the proposed study, the segmentation task was studied. The purpose of this work is to improve the segmentation of overlapping RBCs. To this end, CNN-transformer hybrid architecture, TransUNet was introduced to improve the segmentation with the help of labeled data that promotes the separation of overlapping red blood. The proposed work achieved a 94.5% Jaccard similarity index. This surpassed the results of previous approaches.