Video stitching is crucial in multi-camera-based systems that provide 360-degree surveillance and monitoring applications. Homography estimation is the most important step in video/image stitching. All the existing methods mainly focus on homography estimation through traditional keypoint detection or a more recent deep learning approach. These estimation methods are primarily based on multiple homography calculations and fail for featureless images, which lack keypoints, such as the plain sky. To overcome these limitations, we propose a novel real-time video stitching method based on homography estimation through camera calibration using fixed camera configuration. As the proposed method is based on the relative position of two cameras, the overall image/video stitching process is independent of the world scene. In addition, our method calculates the homography matrix only once during the video stitching process. Thus reducing the per-frame stitching time by about 30% compared to multiple homography estimations. As camera calibration information is unavailable with the existing datasets, we record a novel dataset for benchmarking our method.