Game-Theoretic approach for Dynamic Resource

Media format
Last updated on June 15, 2010
Doug Young Suh
[email protected]
Media signal
Go to basics
“I learned everything in Kindergarten.”
 What did I learn in Kindergarten for video coding?
• Probability theory, linear algebra, sampling theory
AD conversion
Analog = continuous time + analog
Digital = discrete time + digital
Media signal
Sampling and human-being
Sampling theory (Nyquist theory)
Sampling frequency fs > 2fmax
Speech 8kHz > 2 X 3.2kHz
Audio 44.1kHz > 2 X 20kHz
Video
440Hz,
A, la
Temporal : 24Hz (movie), 60Hz (TV)
Spatial
Low freq.
Do-Re-Mi-Pa-Sol-Ra-Si-Do
High frequency
Media signal
Quantization
Analog
digital
Analog : infinite number of values
Digit-al : finite number of values
Media signal
Quantization and de-quantization
 Loss of information, but acceptable accuracy
Media signal
Audio
 Speech: 8kHz sampling Х 8bits/sample = 64kbps
압축  10kbps
 CD quality: 44kHz sampling Х 16bits/sample Х 2(stereo)
1.544Mbps  압축  128kbps or 256kbps MP3
Image Resolution
Spatial resolution: 256*256, 8bits
Temporal resolution: 24Hz, 60Hz, etc.
(fps: frames per second)
128*128, 8bits
256*256, 4bits
Media signal
Color representation
RGB
cf) rectangular coordinate
Red
(255,0,0)
Blue
(0,0,255)
YCbCr (YUV)
cf) polar coordinate
R
Camera
Media signal
G
B
Y
Linear
Matrix
Cb
Cr
A to D
Converter
Color representation
Original
Y
Red
Green
U
V
U, V no high freq.  4:2:2  4:2:0
Blue 4:4:4
Color conversion
 1 to 1 mapping
RGBYCbCr
Y = 0.29900R+0.58700G+0.11400B
Cb = -0.16874R-0.33126G+0.50000B
Cr = 0.50000R-0.41869G-0.08131B
YCbCrRGB
R=1.00000Y+1.40200Cr
G=1.00000Y-0.34414Cb-0.71414Cr
B=1.00000Y+1.77200Cb
Media signal
Color conversion
 4:4:4  1+1+1
4:2:2
Horizontally 2:1
subsampling
4:2:0 1+1/4+1/4
 Most popular
 Amount ½
Media signal
Video formats
 QCIF  CIF(352x288)  4CIF
 QVGAVGA(640x480)SXGA(1280x1024)
Video encoding
2-1
2-3
2-2
Original Video
+
Motion
Estimation
Frame
Memory
Networked Video
DCT
Motion vector
Q
IQ
IDCT
VLC
Encoded
Bitstream