Please enable JavaScript.
Coggle requires JavaScript to display documents.
9_Advanced_Video_Coding_Standards - Coggle Diagram
9_Advanced_Video_Coding_Standards
Motion Compensation
N x N macroblocks
N=16 for luminance
N=8 for chrominance (4:2:0)
macroblock level
image frame
current
target
next
reference
Motion Vector
Optimizing Search
Limited Window
[-p, p]
window size = (2p+1)*(2p+1)
limited neighbourhood
Types
Sequential
MV = (u,v) = [(i,j) | MAD(i,j) min, i∈[-p,p], j∈[-p,p]]
2D Logarithmic
Hierarchical
Mean Absulute Difference (MAD)
next image frame
Supporting Interlaced Videos
Interlaced Video
top-field
bottom-field
Frame-picture
same as non-interlaced
both fields interleaved
single frame
16x16 macroblocks
coded using MC
Field-picture
each field
seperate picture
Modes of Prediction
MPEG-2
Frame Prediction
identical to MPEG1
B-frames
previous & next
fast forward & rewind
better compression
P-frames
previous
I-frames
Zigzag Scanning
Field Prediction
Alternate Scanning
early vertical scan
MPEG-2 Scalability
base layer
lowest quality
resolution
(related)
bits per pixel
Signal to Noise Ratio
color depth
fps
Why?
Networks w/
different bit rates
variable bit rate (VBR) channels
noisy connections
Supports
Data Partitioning
Scability
Temporal
fps
Spatial
resolution
Signal to Noise Ratio (SNR)
bits per pixel
Hybrid
combitnation of above
enhancement layer
high quality - low quality
H.264/MPEG-4 AVC
performance
30% better than H.263
50% better than MPEG-2
upgrades
int DCT 4x4 blocks
MC in luma images
Quarter-pixel accuracy in MV
Multiple reference picture MC
Robust to errors
Motion Compensation
Hybrid Coding
inter-frame
intra-picture
same frame is used for predicting itself
Variable Block-Size Compensation
Quarter-Pixel Precision
Fractional Sample Interpolation
Additional Options in Group of Pictures
Multiple reference frames
Hierarchical prediction structure
No B-frames
