CS4610: L10 - Cameras, Images, Sampling, and Color
- light source, object, (continuous) sensor
- TBD diagram
- also need an aperture or lens
- actual image is effectively continuous in space, time, and spectrum
- need to discretize each of those three with sampling
- pixels sample in space
- frames sample in time
- intensity = num photons/sec
- every photon also has a wavelength
- spectral histogram: photons/sec vs wavelength at a pixel and time
- TBD diagram
- spectral colors: ROYGBIV
- non-spectral colors: brown, pink, white, black, …
- representing color as mixtures of basis histograms
- effectively sampling the space of spectrums
- human eye: rods and s/m/l cones b/g/r
- RGB color space as a cube (R+G = Y, R+B = M, G+B = C)
- TBD diagram
- but aparrent RGB color depends on light intensity, shadows, …
- for machine vision can be advantageous to “factor out” changes in lighting
- tip RGB cube so W is on top, K on bottom
- stretch into “hexcone” with W in center at top
- “hue” = angle, “saturation” = radius, “value” = elevation
- TBD figure
- TBD RGB->HSV equations