Need to align images from different sources? This survey organizes and synthesizes the broad range of image registration techniques used to match two or more pictures taken at different times, from different sensors, or viewpoints. Registration is a fundamental task in image processing required by virtually all large systems that evaluate images. The paper establishes a relationship between the variations in the images and the type of registration techniques that can most appropriately be applied. The study distinguishes three major types of variations affecting image registration. It addresses variations caused by differences in acquisition, which can be removed with spatial transformations. It also considers those variations that are difficult to model, such as lighting conditions and atmospheric effects. Finally, the paper analyzes differences in images that are of interest, such as object movements or growths, which should not be removed by registration. This framework provides a valuable resource for understanding the merits and relationships between various existing techniques and helps select the most suitable technique for a specific problem. The survey is beneficial for researchers and practitioners in fields such as computer vision, remote sensing, and medical imaging.
As a survey, this article is a strong fit for ACM Computing Surveys, which aims to provide comprehensive and structured overviews of important topics in computer science. By organizing and synthesizing the diverse techniques in image registration, the paper offers significant value to the journal’s readership, enabling them to better understand and apply these methods.