Understanding the world through motion: This survey paper provides a comprehensive overview of optical flow, the pattern of apparent motion of objects and surfaces in a visual scene caused by the relative motion between an observer and the scene. Sequences of time-ordered images allow the estimation of projected two-dimensional image motion as either instantaneous image velocities or discrete image displacements. The paper investigates widely known methods for estimating optical flow, classifying and examining them based on the underlying hypotheses and assumptions. It explores how optical flow can be used to recover three-dimensional motion and surface structure, as well as to perform motion detection, object segmentation, time-to-collision calculations, motion-compensated encoding, and stereo disparity measurement. This survey concludes by discussing current research challenges and directions in optical flow computation. It serves as a valuable resource for researchers and practitioners in computer vision, image processing, and related fields.
Published in ACM Computing Surveys, this paper is perfectly aligned with the journal's focus on providing comprehensive overviews of important topics in computer science. By surveying the various methods for computing optical flow and their underlying assumptions, the paper offers a valuable resource for researchers and practitioners in computer vision and related fields.