Image stitching is a method of producing a wider field of view by combining several overlapping images. With four main stages in the image stitching process, the algorithms used at each stage can have a dramatic impact on the success of stitching an image. For each stage, there are a wide range of algorithms to choose from and it can be a challenge to identify a stitching pipeline that will produce the best results. In this paper, we study the approaches involved in each of the four stages of image stitching. A real-world dataset is utilised to evaluate each algorithm, where images are transformed to different perspectives. The similarities of these images are compared to a warped perspective image obtained using the homographies provided by the dataset. The pipelines tested were limited to producing accurate results up to and including a 50° perspective change. Pipelines utilising BRISK’s feature detector, FREAK, and Brute Force produced significant results. However, pipelines incorporating ORB, FAST, or BRIEF produce poor results when compared to other feature detection and feature description algorithms. Generally, the ratio test hindered the matched pairs process, although there were exceptions. Finally, the inlier/outlier detection algorithms, USAC and RANSAC, had similar performances with no definitive data to suggest that, in general, one outperforms the other.