2026-09-19T02:26:00Zhttps://dspace.mit.edu/server/oai/request

oai:dspace.mit.edu:1721.1/1139782026-06-17T14:43:22Zcom_1721.1_7582com_1721.1_7581col_1721.1_131022

William T. Freeman. Xue, Tianfan Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science 2018-03-02T22:21:27Z 2018-03-02T22:21:27Z 2017 2017 http://hdl.handle.net/1721.1/113978 1023628250 Thesis: Ph. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2017. Cataloged from PDF version of thesis. Includes bibliographical references (pages 115-126). Motion is important for understanding our visual world. The human visual system relies heavily on motion perception to recognize the movement of objects, to infer the 3D geometry of a scene, and to perceive the emotions of other people. Modern computer vision systems also use motion signals extracted from video sequences to infer high-level visual concepts, including human activities and abnormal events. Both human and computer visual systems try to perceive changes in the 3D physical world through its 2D projection, either on the image plane or on our retinas. The observed 2D pixel movement is the result of several factors. First, the image sensor might move, inducing egocentric motion, even when the scene is static. Second, the medium between objects and a camera might change and affect how light transmits from the objects to the sensor, like the shimmering in a hot-road mirage. Finally, the objects in a scene might move, either actively, like a person walking along a street, or passively, like a tree branch that is vibrating due to wind. All of these movements reveal information about our visual world. In this dissertation, we will discuss how to infer physical properties of our visual world from observed 2D movement. First, we show how to infer the depth of a scene from egocentric motion and use this to remove undesired visual obstructions. Second, we relate the slight wiggling motion due to refraction to the movement of hot air and infer the location and velocity of the airflow. Last, we illustrate how to infer the physical properties of objects, such as their deformation space or internal structure, from their motion. by Tianfan Xue. Ph.D. 22, 126 pages eng Massachusetts Institute of Technology MIT theses are protected by copyright. They may be viewed, downloaded, or printed from this source but further reproduction or distribution in any format is prohibited without written permission. http://dspace.mit.edu/handle/1721.1/7582 Electrical Engineering and Computer Science. Exploiting visual motion to understand our visual world Thesis application/pdf Publication unknown unknown <Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="7936383b-d632-4ef8-93b8-790cd9ef0044"> <Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type> <Language>eng</Language> <Title>Exploiting visual motion to understand our visual world</Title> <PublishedIn> <Publication> </Publication> </PublishedIn> <PublicationDate>2017</PublicationDate> <Authors> <Author> <DisplayName>Xue, Tianfan</DisplayName> <Affiliation> <OrgUnit> </OrgUnit> </Affiliation> </Author> </Authors> <Editors> </Editors> <Publishers> <Publisher> <DisplayName>Massachusetts Institute of Technology</DisplayName> <OrgUnit /> </Publisher> </Publishers> <License>http://dspace.mit.edu/handle/1721.1/7582</License> <Keyword>Electrical Engineering and Computer Science.</Keyword> <Abstract>Motion is important for understanding our visual world. The human visual system relies heavily on motion perception to recognize the movement of objects, to infer the 3D geometry of a scene, and to perceive the emotions of other people. Modern computer vision systems also use motion signals extracted from video sequences to infer high-level visual concepts, including human activities and abnormal events. Both human and computer visual systems try to perceive changes in the 3D physical world through its 2D projection, either on the image plane or on our retinas. The observed 2D pixel movement is the result of several factors. First, the image sensor might move, inducing egocentric motion, even when the scene is static. Second, the medium between objects and a camera might change and affect how light transmits from the objects to the sensor, like the shimmering in a hot-road mirage. Finally, the objects in a scene might move, either actively, like a person walking along a street, or passively, like a tree branch that is vibrating due to wind. All of these movements reveal information about our visual world. In this dissertation, we will discuss how to infer physical properties of our visual world from observed 2D movement. First, we show how to infer the depth of a scene from egocentric motion and use this to remove undesired visual obstructions. Second, we relate the slight wiggling motion due to refraction to the movement of hot air and infer the location and velocity of the airflow. Last, we illustrate how to infer the physical properties of objects, such as their deformation space or internal structure, from their motion.</Abstract> <Access xmlns="http://purl.org/coar/access_right" > </Access> </Publication>