Stefano Soatto (c)UCLA Vision Lab5What is vision? From the 3-D world to 2-D images: imageformation (physics). Domain of artistic reproduction (synthesis):painting, graphics. From 2-D images to the 3-D world: imageanalysis (mathematical modeling, inference). Domain of vision: biological (eye+brain),computationalStefano Soatto (c)UCLA Vision Lab6[Felleman & Van Essen, 1991] This is the partof your brainthat processesvisualinformationVISUAL INFORMATIONPROCESSINGStefano Soatto (c)UCLA Vision Lab7This is a picture of my officeThis is how a computerrepresents itStefano Soatto (c)UCLA Vision Lab8This is a picture of my officeAnd so is this …Stefano Soatto (c)UCLA Vision Lab9This is a picture of my officeAnd so arethese!We need to extract some “invariant”, i.e. what is common toall these images (they are all images of my office)Stefano Soatto (c)UCLA Vision Lab12SIDE EFFECTS OF LATERALINHIBITIONTHESE ARE NOT SPIRALSTHESE ARE NOT SPIRALSAND THEY ARE NOT MOVING!AND THEY ARE NOT MOVING!Stefano Soatto (c)UCLA Vision Lab13IMAGE SYNTHESIS: simulation of the image-formation process • Pinhole (perspective) imaging in most ancient civilizations.• Euclid, perspective projection, 4th century B.C., Alexandria (Egypt)• Pompeii frescos, 1st century A.D. (ubiquitous).• Geometry understood very early on, then forgotten.Image courtesy of C. TaylorStefano Soatto (c)UCLA Vision Lab14PERSPECTIVE IMAGING (geometry) Image courtesy of C. Taylor• Re-discovered and formalized in the Renaissance:• Fillippo Brunelleschi, first Renaissance artist to paint with correct perspective,1413• “Della Pictura”, Leon Battista Alberti, 1435, first treatise• Leonardo Da Vinci, stereopsis, shading, color, 1500s• Raphael, 1518Stefano Soatto (c)UCLA Vision Lab15IMAGE ANALYSIS: THE INVERSE PROBLEM Input: Images (measurements of LIGHT)Intermediate representation: “Features” (2-D geometry)Output: Camera calibration, 3-D pose, scene structure, surface photometry.IN THIS CLASS: only geometry; for photometry take CS174BStefano Soatto (c)UCLA Vision Lab16IMAGES AND GEOMETRY – History of “Modern” Geometric Vision • Chasles, formulated the two-view seven-point problem in a class homework assignment in 1855• Hesse, solved the above problem, 1863• Kruppa, solved the two-view five-point problem, 1913• Longuet-Higgins, the two-view eight-point algorithm, 1981• Liu and Huang, the three-view trilinear constraints, 1986• Faugeras, uncalibrated reconstruction, 1992• Tomasi and Kanade, (orthographic) factorization method, 1992• iata, iata, iata …• MaSKS: generalized rank conditions, 2003.Stefano Soatto (c)UCLA Vision Lab17APPLICATIONS – 3-D Modeling and RenderingStefano Soatto (c)UCLA Vision Lab18APPLICATIONS – 3-D Modeling and Rendering Image courtesy of Paul DebevecStefano Soatto (c)UCLA Vision Lab19APPLICATIONS – Image Morphing, Mosaicing, Alignment Images of CSL, UIUCStefano Soatto (c)UCLA Vision Lab203D RECONSTRUCTION FROM3D RECONSTRUCTION FROMMULTIPLE VIEWS:MULTIPLE VIEWS:GEOMETRY AND GEOMETRY AND PHOTOMETRyPHOTOMETRywith h. jin; image courtesy: j-y bouguet - intelStefano Soatto (c)UCLA Vision Lab21estimated shapelaser-scanned,manually polishedwith h. jinStefano Soatto (c)UCLA Vision Lab22estimate both geometry estimate both geometry andand photometry photometrywith h. jinStefano Soatto (c)UCLA Vision Lab23with h. jinStefano Soatto (c)UCLA Vision Lab24image courtesy: j-y bouguet - intelStefano Soatto (c)UCLA Vision Lab25Stefano Soatto (c)UCLA Vision Lab26APPLICATIONS – Real-Time Sports Coverage Image courtesy of Princeton Video Image, Inc.First-down line and virtual advertisingStefano Soatto (c)UCLA Vision Lab27APPLICATIONS – Real-Time Virtual Object Insertion UCLA Vision LabStefano Soatto (c)UCLA Vision Lab28APPLICATIONS – Unmanned Aerial Vehicles (UAVs)Berkeley Aerial Robot (BEAR) ProjectRate: 10HzAccuracy: 5cm, 4oStefano Soatto (c)UCLA Vision Lab29APPLICATIONS – Autonomous Highway Vehicles Image courtesy of E.D.
View Full Document