HOME

IMAGES

FINAL REPORT
DCT Transform
Haar Wavelet Transform

quantization methods
results
conclusions

REFERENCES/LINKS


RESULTS

wpe67.jpg (2127 bytes)

Both the DCT and Haar Wavelet codes were run several times using various quantization schemes and parameters.  Some of the MSE (mean squared error) and PSNR (peak signal to noise ratio) information obtained from these runs is included in tables below.  The compression information for the binary files resulting from these tests is also included.  The quality of the results can be seen in the accompanying images.

 

DCT Results

Quantitative Analysis
Image file: KGirl.256

# of bits MSE PSNR
8 574.6263 20.5369
16 221.2136 24.2827
24 79.5294 29.1255
32 3.5922 42.5771

Compression Analysis

# of bits Compressed (gzip) Ratio
8 2177 30.1:1
16 8865 7.4:1
24 16640 3.9:1
32 24941 2.6:1

DCT Transform Images

Result1.jpg (12624 bytes)

8 bit quantization                                            ;             24 bit quantization

 

Result2.jpg (12957 bytes)

32 bit quantization

 

Haar Wavelet Results

Quantitative Analysis (Levels: 4)

Method Image File # of bits MSE PSNR
Uniform KGirl.256 4 99.36 28.16
Adaptive KGirl.256 8,6,4,2,0 124.21 27.19
Statistical KGirl.256 8,6,4,2,0 60.19 35.73
Adaptive Baby.256 8,6,4,2,0 182.76 25.51
Adaptive Building.256 8,6,4,2,0 84.75 28.85

Compression Analysis (Levels: 4)

Method Image File # of bits Compressed (gzip) Ratio
Uniform KGirl.256 4 32768 7:1
Adaptive KGirl.256 8,6,4,2,0 5440 33.7:1
Statictical KGirl.256 8,6,4,2,0 5440 35.1:1
Adaptive Baby.256 8,6,4,2,0 5440 39.5:1
Adaptive Building.256 8,6,4,2,0 5440 33.6:1

 

Haar Transform Images

wpe18.jpg (7791 bytes)

Uniform Quantization, Levels: 4, Bits:  4

 

wpe19.jpg (7181 bytes)

Adaptive Quantization, Levels: 4, Bits: [8 6 4 2 0]

 

s_l4_b86420.GIF (176527 bytes)

Statistical Quantization, Levels: 4, Bits: [8 6 4 2 0]

wpe67.jpg (2127 bytes)

Contact the authors:
nofences@rice.edu
heidit@caam.rice.edu
hard@rice.edu
bwang@rice.edu