CS248 Midterm Review

Michael Green and Sean Walker

(based on the work of previous TAs)

CS248 MidtermCS248 Midterm

• Mon, November 1, 7-9 pm, Gates B01 • Mostly “short answer” questions

– Keep your answers short and sweet!

• Covers lectures up to Tuesday, Oct 26– plus taxonomy from start of last lecture

• Review session slides available from class website

• Exam is closed book, closed notes

Raster Displays, Raster Displays, Resolution, PerceptionResolution, Perception

• CRTs– 3 phosphors for “red”, “green”, and

“blue”– Triads and shadow mask

• Measures of spatial resolution– physical vs. addressable resolution

Human spatial frequency Human spatial frequency sensitivitysensitivity

– Sensitivity highest in fovea

– Frequency sensitivity– Phase sensitivity (Vernier acuity)– Temporal sensitivity

• Flicker (50-70Hz)• Perceived motion

– 12 Hz = cartoons, 24 Hz = film, 60 Hz = video

Raster Displays, Raster Displays, Resolution, PerceptionResolution, Perception

• Human intensity sensitivity– Response to intensity is nonlinear– Gamma in cameras, CRTs – Gamma correction

Raster Displays, Raster Displays, Resolution, PerceptionResolution, Perception

• Sample (easy) question:1. A scene is photographed with a TV

camera with gamma=0.5 and displayed on a CRT with gamma=2.4. If we want system gamma to be 1.0, we should do gamma correction with what exponent?

ColorColor

• Perception of color– Humans are trichromat

• Three cones sensitive to “red”, “green”, and “blue”

– Overlapping response curves• Know their general shapes!

• Color matching– Color matching experiment

Color spacesColor spaces

• Linear colorspaces– , , space (perceptual stimulus)– R, G, B space – X, Y, Z space

• Non-linear colorspaces– HSV

• Spectral locus• Gamut of reproducible colors

ColorColor

Sample questions:1. Circle those colors that are not

reproducible with a single monochromatic light1. Red2. Yellow3. Blue4. Magenta5. White6. Green

ColorColor

Sample questions:2. If you had a special CRT that could produce pure spectral colors, how many spectral colors would you need to represent a normal RGB color gamut? How about the spectral locus?

Sampling and AntialiasingSampling and Antialiasing

• The sampling and reconstruction pipeline:– Prefiltering– Sampling– Resampling– Reconstruction

• Aliasing in the frequency domain• Filtering and convolution

– Duality: F(x)*G(x) <=> F(w)G(w)

Sampling and AntialiasingSampling and Antialiasing

• Prefiltering vs. postfiltering• Desirable filters for antialiasing

– Box, pyramid, gaussian, sinc

• Methods of antialiasing– Supersampling: regular vs. stochastic– Analytical antialiasing

Sampling and AntialiasingSampling and Antialiasing

Sample questions:1. What is the result of convolving a 1-

D box filter with itself?

2. Which of the following would affect your choice of a reconstruction filter?a) pixel shapeb) choice of prefilterc) actual size of display

RasterizationRasterization

• Rasterization of lines– DDA, incremental algorithm

• Rasterization of polygons– Only pixels in the polygon

• Supersampling– Patterns: understand its effect on the

image

RasterizationRasterization

• Sample question:– If you rasterized this line using DDA,

which pixels would get turned on?

Digital CompositingDigital Compositing

• What is compositing?• The compositing approximation

– Conditions for validity

Compositing algebraCompositing algebra

Digital CompositingDigital Compositing

Sample question:You are doing the special effects for a

movie, and need to composite a computer generated object over a live background. Why should you use an 8-bit alpha matte rather than a binary (1-bit) matte, even if the computer-generated object is fully opaque?

TransformationsTransformations

• Homogeneous coordinates – why?• Matrices rotation, translation, scale, shear in 2D,

3D

– Know the form of each kind– Geometric properties preserved/changed by

each kind

• Composing transformations – multiply matrices in reverse order

TransformationsTransformations

Sample questions• Compute the 2D transform that translates an

object centered at (-3,4) to the origin, then rotates it by +45o, then translates it to (10,5).

• What sequence of transforms would cause the triangle to change as shown below ?

GOOD LUCKANDHAPPY HALLOWEEN!