Image segmentation is considered as a fundamental topic in computer vision and image processing. The primary aim of segmentation is to partition an image into parts that are uniform in nature, and enables the subsequent analysis. It plays a crucial role in the field of medical imaging for disease diagnosis. There are many technologies available for medical image segmentation, and among them deep learning (DL) methods are highly implemented due to their feature extraction characteristic. Thus, in the presented work, an ensemble model is proposed to perform the binary segmentation of medical images. The proposed model is formulated from two base models that are combined by stacking approach of ensemble learning through a meta-learner model. The base models are inspired from the U-Net and SegNet architecture, and the meta-learner model is designed from convolutional neural network (CNN) architecture. The study presents an analysis to evaluate the proposed model performance on three distinct types of medical images. The medical images utilized are breast ultrasound images, gastrointestinal (GI) tract images and chest X-ray images. The performance of ensemble model is assessed by segmented masks and various performance measures like, accuracy, recall, precision, and mean intersection over union (mIoU). The experiments showcase that the proposed ensemble model performs comparatively better than the base models. Moreover, the ensemble model yields highest accuracy and mIoU value as 0.9076 and 0.6939, respectively, corresponding to GI tract images.