The following are the numerical comparisons on some 512x512 grayscale images. The bit rates are calculated from the actual file sizes (note that by adjusting the quantizer step size the target bit rate can be exactly matched. These bit rates also include about a 10 byte header containing the image dimensions, number of wavelet scales, quantizer step size, and area threshold.

Algorithm/Rate [bpp]
0.125
0.25
0.30
0.40
0.50
1.00
EZW [1]
30.23
33.17
-
-
36.28
39.55
MRWD [2]
-
-
34.07
-
36.60
40.17
MRWD [5]
31.09
34.12
34.93
36.20
37.17
40.33
SPIHT [3]
31.09
34.11
34.95
36.24
37.21
40.41
SLCCA1 [4]
31.38
34.33
35.14
36.42
37.38
40.44
SLCCA2 [4]
31.25
34.28
35.08
36.37
37.35
40.47
Table 1. Performance comparison (PSNR [dB]) on 512x512 "Lena" image.

Algorithm/Rate [bpp]
0.125
0.25
0.30
0.40
0.50
1.00
EZW [1]
24.03
26.77
-
-
30.53
35.14
MRWD [5]
25.27
27.86
28.71
30.22
31.44
36.24
SPIHT [3]
24.86
27.58
28.56
30.10
31.39
36.41
SLCCA1 [4]
25.40
28.39
29.35
30.89
32.24
37.12
SLCCA2[4]
25.36
28.18
29.05
30.55
31.89
36.69
Table 2. Performance comparison (PSNR [dB]) on 512x512 "Barbara" image.

Rate [bpp] Algorithm
Image
"Baboon"
"Couple"
"Man"
"Boat"
"Tank"
"Goldhill"
0.25 MRWD [5]
23.26
29.19
29.95
30.92
29.42
30.53
0.25 SPIHT [3]
23.27
29.25
30.01
30.97
29.36
30.56
0.25 SLCCA1 [4]
23.44
29.37
30.07
31.09
29.44
30.60
0.25 SLCCA2 [4]
23.45
29.35
30.10
31.11
29.44
30.65
0.30 MRWD [5]
23.83
29.93
30.69
31.77
29.86
31.14
0.30 SPIHT [3]
23.76
30.07
30.74
31.77
29.77
31.15
0.30 SLCCA1 [4]
23.99
30.09
30.81
31.94
29.89
31.22
0.30 SLCCA2 [4]
23.98
30.12
30.83
31.96
29.89
31.27
0.40 MRWD [5]
24.84
31.25
31.95
33.20
30.61
32.19
0.40 SPIHT [3]
24.66
31.29
31.94
33.16
30.52
32.18
0.40 SLCCA1 [4]
24.96
31.42
32.04
33.44
30.65
32.29
0.40 SLCCA2 [4]
24.96
31.42
32.08
33.39
30.62
33.36
0.50 MRWD [5]
25.74
32.36
33.06
34.39
31.26
33.15
0.50 SPIHT [3]
25.64
32.45
33.08
34.45
31.16
33.13
0.50 SLCCA1 [4]
25.83
32.54
33.12
34.68
31.28
33.24
0.50 SLCCA2 [4]
25.86
32.59
33.19
34.61
31.27
33.26
1.00 MRWD [5]
29.35
36.48
37.26
38.96
33.99
36.56
1.00 SPIHT [3]
29.17
36.58
37.34
39.12
33.78
36.55
1.00 SLCCA1 [4]
29.33
36.56
37.32
39.28
33.99
36.63
1.00 SLCCA2 [4]
29.36
36.64
37.45
39.24
34.04
36.66
Table 3. Performance comparison (PSNR [dB]) of SPIHT and SLCCA on different 512x512 natural images.

Image
SLCCA1 [4]
SLCCA2 [4]
SPIHT [3]
fingerprint
27.61
27.37
27.07
sweater
41.83
41.84
41.48
grass
25.45
25.33
24.82
pig skin
26.82
26.69
26.50
raffia
20.93
20.93
20.30
sand
24.18
24.10
23.63
water
29.76
29.35
29.19
wool
26.40
26.14
25.70
Table 4. Performance comparison (PSNR [dB]) of SPIHT and SLCCA on 256x256 texture images at 0.4 bpp.

Algorithm/Rate [bpp]
0.125
0.25
0.30
0.40
0.444
0.5
1.00
WSQ
-
-
-
-
34.43
-
-
SPIHT [3]
29.46
32.82
33.62
34.98
35.45
36.03
40.00
SLCCA1 [4]
29.97
33.29
34.08
35.34
35.81
36.34
40.39
SLCCA2
29.80
33.09
33.89
35.16
35.65
36.25
40.32
Table 5. Performance comparison (PSNR [dB]) on 768x768 "Fingerprint" image.

References

[1] J. Shapiro, "Embedded image coding using zerotrees of wavelet coefficients," IEEE Transactions on Signal Processing, vol. 41, no. 12, pp. 3445-3462, Dec. 1993.

[2] S. Servetto, K. Ramchandran, and M. Orchard, "Wavelet based image coding via morphological prediction of significance," Proceedings of IEEE ICIP-95, Oct. 1995, pp. 530-533.

[3] A. Said and W.A. Pearlman, "A new, fast, and efficient image codec based on set partitioning in hierarchical trees," IEEE Transactions on Circuits and Systems for Video Technology, vol. 6, no. 3, pp. 243-250, June 1996.

[4] B.-B. Chai, J. Vass, and X. Zhuang, "Significance-linked connected component analysis for low bit rate image coding," in Proceedings of IEEE International Conference on Image Processing, Santa Barbara, CA, Oct. 26-29, 1997, pp. 637-640.

SLCCA

1. Villasenor 10/18 filter obtained from ftp.cs.dartmouth.edu

2. Antonini 9/7 filter from M. Antonini, M. Barlaud, P. Mathieu, and I. Daubechies, "Image coding using wavelet transform," IEEE Transactions on Image Processing, vol. 1, no. 2, pp. 205-220, Apr. 1992.