everything.study
6.866

Machine Vision

Undergraduate · Historical course

Fall 2020

Topics

  • Overview and image formation. Where will it image? How bright will it be? How can we invert image formation?
  • Image formation, perspective projection, time derivative, and motion field
  • Time to contact, focus of expansion, direct motion vision methods, and noise gain
  • Fixed optical flow, optical mouse, constant brightness assumption, and closed form solution
  • TTC and FOR Montivision demos, vanishing point, and use of VPs in camera calibration
  • Photometric Stereo, noise Gain, error amplification, eigenvalues, and eigenvectors review
  • Gradient space, reflectance map, image irradiance equation, and gnomonic projection
  • Shape from shading in special cases, lunar surface, scanning electron microscope, and Green’s theorem in photometric stereo
  • Shape from shading in general case—from first order nonlinear PDE to five ODEs
  • Characteristic strip expansion, shape from shading, and iterative solutions
  • Edge detection, subpixel position, CORDIC, and line detection (US Patent 6,408,109)
  • Blob analysis, binary image processing, use of Green’s theorem, and derivative and integral as convolutions
  • Object detection, recognition and pose determination, and PatQuick (US Patent 7,016,539)
  • Inspection in PatQuick, Hough transform, homography, position determination, and multi-scale
  • Alignment, recognition in PatMAx, distance field, and filtering and sub-sampling (US Patent 7,065,262)
  • Fast convolution, low pass filter approximations, and integral images (US Patent 6,457,032)
  • Photogrammetry, orientation, axes of inertia, symmetry, and absolute, relative, interior, and exterior orientation
  • Rotation and how to represent it, unit quaternions, and the space of rotations
  • Absolute orientation in closed form, outliers and robusteness, and RANSAC
  • Space of rotations, regular tesselations, and critical surfaces in motion vision and binocular stereo
  • Relative orientation, binocular stereo, structure from motion, quadrics, camera calibration, and reprojection
  • Exterior orientation, recovering position and orientation, bundle adjustment, and object shape
  • Gaussian image and extended Gaussian image, solids of revolution, direction histograms, and regular polyhedra
  • Putting it all together
Fall 2004

Topics

  • Shape from Shading
  • Shape from Gradient (cont.)
  • Binary Image Processing (cont.)
  • The Euler Equations and their Applications
  • Optical Flow
  • Optical Flow (cont.)

Readings

Sources