A DCTDomain Video Transcoder for Spatial Resolution Downconversion

1
A DCT-Domain Video Transcoder for Spatial
Resolution Downconversion

• Yuh-Reuy Lee, Chia-Wen Lin and Cheng-Chien Kao
• CS_at_National Chung Cheng University
• Presented by Bin Yu

2
Outline

• Motivation
• Transcoder Architecture
• Spatial Downscaling
• Sub-sampling
• DCT Decimation
• Motion Vector Composition
• Conclusion

3
Motivation

• Universal Multimedia Access
• Heterogeneous devices/connections in the network

Spatial Downscaling
4
Spatial Resolution Downscaling Illustration
4
6
8
3
3
6
4
3
2
1
3
-1
2
5
9
6
7
8
8
9
6
5
2
7
4
2
5
1
10
5
6
3
9
8
9
11
11
9
10
9


7
8
2
6
8
6
4
2
4
3
16
18
5
14
8
4
6
4
2
4
7
9
1
3
5
7
10
12
4
7
7
5
3
5
7
5
2
4
3
2
Motion Vectors and Differences
Resulting Macroblocks
4
6
8
3
3
6
4
3
2
5
9
6
7
8
2
7
Reference Macroblocks
Downscaling
6
7
4
4
Point to ?
?
4
5
8
8
9
8
2
2


5
5
4
4
14
8
6
3
6
7
4
4
5
Cascaded DCT-domain Transcoder Architecture
6
Cascaded DCT-domain Spatial Resolution
Downscaling Transcoder Architecture
7
Spatial Downscaling scheme 1 Sub-sampling
Pixel Domain Sub-sampling
DCT Domain Equivalent
b1 8x8
b2 8x8
b1 4x4
b2 4x4
b3 4x4
b4 4x4
b3 8x8
b4 8x8
b
Example
h1
g1 h1
8
Spatial Downscaling scheme 2 DCT Decimation

• Key observation hi are sparse, but DCT(hi) are
  not!

8x8 DCT
B1 8x8
B2 8x8
4x4 IDCT
B3 8x8
B4 8x8
9
Motion Vector Composition

• Composing one new MV from 4 old MVs
• Majority voting
• Averaging
• Median filtering the one MV has the least
  distance to all other MVs
• Activity-weighted Median
• Use squared sum of all coefficients as the
  activity of a MB

10
Conclusion

• Contribution
• A new dimension of stream adaptation
• A suite of DCT-compressed domain algorithms
• Future work
• Real-world implementation
• Going for HDTV
• HDTV 1920x1088
• PC/Laptop 960x544
• iPAQ 240x136