Daniel Suo

Scientific progress goes 'boink'

Ph.D. Candidate
Princeton University
Department of Computer Science

Computer Vision

These notes are based on the course COS429, which is taught at Princeton by Professor Jianxiong Xiao.

Notes with their own pages

Notes that have not been organized

Images

Pixels

From class notes: Conceptually, each pixel is a sample of the radiance arriving at a camera viewpoint from a direction. The camera records the number of photons captured for each pixel in its sensor in a given period of time and approximates the power (energy from the photons in a given time) for all pixels.

Photograph

From class notes: A photograph is a slice of the plenoptic function

  • At a particular point of view $(x,y,z)$
  • In the camera’s field of view (range over $\theta$ and $\phi$)
  • At a certain time $t$
  • Over a range of frequencies $\lambda$

References

Image operations

Image transformations

  • Image warping changes the domain of the image. In other words, we move pixels around (e.g., moving, scaling, rotating, fisheye).
  • Image filtering changes the range of the image. In other words, we modify the pixels, but keep them in the same location (e.g., Instagram filters).

Fourier transforms

References

vision

3d-rotations

camera-matrix

color

homogenous-coordinates

homogenous-coordinates

index

large-small-aperture

lens-center-rays

lens-equation

lens-normal-rays

lens-trace-rays

lighting

ols

optics

pinhole-2d-geometry

pinhole-3d-geometry

pinhole-no-hole

pinhole-with-hole

pinhole-with-lens

projections

transformations

transformations