A ZeroValue Prediction Technique for Fast DCT Computation

1
A Zero-Value Prediction Technique for Fast DCT
Computation

• Y. Nishida, K. Inoue, and V. Moshnyaga

2
Problem

• Discrete Cosine Transform (DCT) is a very
• computationally expensive task.

• DCT requires a measurable amount of energy
• and many operations.

Reducing the number of operations and energy
consumption of DCT becomes important.
3
Solution
A Zero-Value Prediction Technique for Fast DCT
Computation.
When the DCT outputs N zeros, we predict the
remaining result is also zero.
Example of N2
The detected string of zeros
The predicted values
Starting point of zero-value prediction
4
Advantage and Disadvantage
Advantage

• Reduces the number of DCT and quantization
  
• operations.
• Reduces energy consumption.
• DCT speed increases.

Disadvantage

• The picture quality deteriorates a little.

5
Operation reduction rate as a function of N(DCT)
?
?
?
carphone
?
carphone
foreman
missa
salesman
foreman
missa
salesman
?

?

Operation reduction rate
Operation reduction rate
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
N
N

• The reduction rate is high when N is small.

6
PSNR as a function of N
70
60
50
PSNR (dB)
40
30
20
10
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
N

• The PSNR is low when N is small.
• When N9, can reduce the number of DCT
  operations by 29.
• The PSNR decreases by 1.47dB.