JPEG 2000 Image Analysis

1
JPEG 2000 Image Analysis

• Darius Fennell
• University of Rochester

2
Experiment Objective

• Using JPEG 2000 calculate and analyze the
  following
• Distortion vs. Rate
• Distortion vs. Quantization
• Time vs. Quantization
• Time vs. Resolution

3
Experiment Materials

• Dell Computer
• 2.79 GHz Processing
• 2.00 GB of RAM
• Matlab 7.0.1 R14
• Kakadu JPEG 2000 software
• 5 Standard images

4
Standard Imagery

• Originally 3 bands _at_ 512x512
• Reduced to single band 128x128

5
Basic JPEG 2000 Compression Chain
6
Expectation

• Distortion vs. Rate
• As rate increases distortion decreases
• Distortion vs. Quantization
• As step size decreases distortion decreases
• Time vs. Quantization
• As quantization increases time decreases
• Time vs. Spatial Resolution
• As Spatial Resolution increases time increases

7
Distortion vs. Rate

• Imagery compressed at multiple rates
• 7.0,4.0,2.0,1.0,0.5, and 0.25
• Average Bit Rate per pixel
• Imagery decompressed to each rate and compared to
  the original.
• MSEvar(original-decompressed)

8
Distortion vs. Rate Results
9
Quantization

• Imagery compressed using varying step sizes2-7
  to 27
• MSE calculated
• Processing time determined by Matlabs tic/toc
  command

10
Distortion vs. Quantization Results
11
Time vs. Quantization Results
12
Time vs. Spatial Resolution

• Images Sizes
• 512x512
• 256x256
• 128x128
• 64x64
• Processing time determined by Matlabs tic/toc
  command

13
Time vs. Spatial Resolution
14
Conclusion

• Rate quantization have a drastic effect on the
  error.
• Step size does not significantly effect
  compression time for quantization.
• Increased spatial resolution increases
  compression time yet other factors are involved
  that affect the over processing chain.

15
References

• http//www.kakadusoftware.com
• http//www.gvsu.edu/math/wavelets/student_work/EF/
  index.html