Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
3 2 1 0
3 2 1 0
3 2 1 0
1 3 2 1 0
G G G G 0 0
H H H H 0 0
0 ... 0 G G G G
0 0 H H H H
Where H
i
and Gi are low and high pass filter impulse responses, are 2-by-2 matrices which can be
written as follows:
(
(
(
(
(
(
(
(
(
(
(
... G G G G 0 0 0 0
.... G G G G 0 0 0 0
.... H H H H 0 0 0 0
... H H H H 0 0 0 0
... ... ... ... ... G G G G
... ... ... ... ... G G G G
... ... ... ... ... H H H H
... ... ... ... ... H H H H
1 1 0 0
1 1 0 0
1 1 0 0
1 1 0 0
1 1 0 0
1 1 0 0
1 1 0 0
1 1 0 0
By examining the transform matrices of the scalar wavelet and multi-wavelets, it is observed
that in multi-wavelets transform domain there are first and second low-pass coefficients followed by
first and second high pass filter coefficients rather than one lowpass coefficient followed by one high
pass coefficient. Therefore, if we separate these four coefficients, there are four sub bands in the
transform domain. Since multi-wavelet decompositions produce two low-pass sub bands and two high
pass sub bands in each dimension, the organization and statistics of multiwavelet sub band differ from
the scalar wavelet case.
Figure 5: Image decomposition after a single level scaling for (a) Scalar wavelets and (b) multi-
wavelets.
.
September Issue Page 64 of 105 ISSN 2229 5216
International Journal of Advances in Science and Technology,
Vol. 3, No.3, 2011
Figure 6: Conventional iteration of multiwavelet decomposition.
During a single level of decomposition using a scalar wavelet transform, the 2- D medical image
data is replaced by four blocks corresponding to the sub bands representing either low pass or high pass
in both dimensions. These sub bands are illustrated in Fig. 5. The sub band labels in Fig. 6 indicate
how the sub band data were generated. For example, the data in sub band LH was obtained from high
pass filtering of the rows and then low pass filtering of the columns. The multi-wavelets used here have
two channels, so there will be two sets of scaling coefficients and two sets of wavelet coefficients.
Since multiple iteration over the low pass data is desired, the scaling coefficients for the two channels
are stored together. Likewise, the wavelet coefficients for the two channels are also stored together.
The multi-wavelet decomposition sub bands are shown in Fig.6. For multi-wavelets the L and H have
subscripts denoting the channel to which the data corresponds. For example, the sub band labeled
L1H2 corresponds to data from the second channel high pass filter in the horizontal direction and the
first channel low pass filter in the vertical direction. This shows how a single level of decomposition is
done. In practice, there is more than one decomposition performed on the medical image. Successive
iterations are performed on the low pass coefficients from the pervious stage to further reduce the
number of low pass coefficients. Since the low pass coefficients contain most of the original signal
Figure 6: Conventional iteration of multiwavelet decomposition., this iteration process yields better
energy compaction. After a certain number of iterations, the benefits gained in energy compaction
becomes rather negligible compared to the extra computational effort. Usually five levels of
decomposition are used. A single level of decomposition with a symmetric-antisymmetric multi-
wavelet is roughly equivalent to two levels of wavelet decomposition. Thus a 3level multiwavelet
decomposition effectively corresponds to 6- level scalar wavelet decomposition.
The scalar wavelet transform gives a single quarter-sized sub band from the original larger sub
band, as shown in fig 1. The multi-level decomposition is performed in the same way. The multi-
wavelet decomposition iterates on the low pass coefficients from the pervious decomposition, as shown
in Fig. 2. In the case of the scalar wavelets, the low pass quarter image is a single sub band. But when
the multi-wavelet transform is used, the quarter image of low pass coefficients is actually a 2 x 2 block
of sub bands (the L_Lj sub bands in Fig. 6. Due to the nature of the preprocessing and symmetric
extension method, data in these different sub bands becomes intermixed during iteration of the
multiwavelet transform. The intermixing of the multiwavelet low pass sub bands leads to suboptimal
results. Consider the multi-wavelets transform coefficients resulting from single-level decomposition.
It can be readily observed that the 2 x 2 "low pass" block (upper left corner) actually contains one low
pass sub band and three band pass sub bands.
The L1L1 sub band resembles a smaller version of the original image, which is a typical
characteristic of a true low pass sub band. In contrast, the L1L2, L2L1, and L2L2 sub bands seem to
process characteristics more like those of high sub bands. Also only L1L1 sub band contains
September Issue Page 65 of 105 ISSN 2229 5216
International Journal of Advances in Science and Technology,
Vol. 3, No.3, 2011
coefficients with a large DC value and a relatively uniform distribution. The L1, L1, H1 and H2 sub
bands, measured along the vertical direction. These description resolution in multiwavelet features are
considered as textural information and are used for image representation for image retrieval systems.
The proposed retrieval system is presented in next section.
4. System architecture
Figure 7:Block diagram for the proposed Hierarchal coding system
The Hierarchal Zero Wavelet output stream starts with information to synchronize the
decoder. The minimum information required by the decoder for its functionality is the number of
wavelet transforms levels used and the initial limit, basically a constant levels (3) of wavelet transform
were used for transformation. The first step in the Hierarchal Zero Wavelet coding algorithm is to
determine the initial limit. The initial limit t
0
is given as
( ) ( )
) y , x (max
2
log
0
2 t
= 3
where MAX(|
T
(x,y)|) means the maximum coefficient value in the image and (x,y) denotes the
coefficient. Then taking the obtained limit as the initial value the scaled sub-band samples are been
runed for initial run and intermediate run. Under each run the limit is decreased by half the value. This
comparison is carried out until the limit reaches to the minimum limit, the algorithm implemented is;
limit = initial_limit;
do {
initial_run(image);
intermediate_run(image);
limit = limit/2;
} while (limit > minimum_limit);
(A) Initial run
The image is scanned and a symbol is returned for every coefficient. If the coefficient is
larger than the limit a P is coded.
1. If the coefficient is smaller than negative of limit an N is coded.
2. If the coefficient is the root of a zero tree then a T is coded and finally,
3. If the coefficient is smaller than the limit but it is not the root of a zero tree, then a Z is coded. This
happens when the coefficient larger than the limit in the sub tree. Finally, all the coefficients that are
in positive value, larger than the current limit are extracted and placed without their sign on the
intermediate list and their positions in the image are filled with zeroes. This prevents them from
being coded again.
September Issue Page 66 of 105 ISSN 2229 5216
International Journal of Advances in Science and Technology,
Vol. 3, No.3, 2011
The initial run can thus be implemented as;
* Initial run
initialize_fifo();
while (fifo_not_empty)
{
get_coded_coefficient_from_fifo();
if coefficient was coded as P, N or Z then
{
code_next_scan_coefficient();
put_coded_coefficient_in_fifo();
if coefficient was coded as P or N then
{
add abs(coefficient) to intermediate list;
set coefficient position to zero;
}
}
}
After the initial run follows the intermediate run:
Intermediate run
intermediate_limit = current_limit/2;
for all elements on intermediate list do
{
if (coefficient > intermediate_limit)
{
output a one;
coefficient=coefficient intermediate_limit;
}
else output a zero;
}
If the limit used is a power of two, then the intermediate run reduces to a few logical
operations and can be made very fast. The decoding unit reconstructs the values by identifying the
symbols as positive, negative, zero tree and isolated zerotree. The reconstructed values are taken as
limit for positive coded coefficients and negative of limit for negative coded coefficients. The zerotree
coefficients and the isolated zerotree coefficients are assigned with 0 value.
5. Results observations
Input properties of more than 150 images are carried out and few cases are represented in table 1. The
corresponding images are also shown in fig 8. and Fig 9. Plot of PSNR comparison between wavelets
and multiwavelets is given in figure 10.
September Issue Page 67 of 105 ISSN 2229 5216
International Journal of Advances in Science and Technology,
Vol. 3, No.3, 2011
Table 1. input properties of different two images
Figure 8. (A) Original image (B) Wavelet based output (C) multiwavelet output
Image
Name
Dimension
Bit
depth
Wavelet Based Compression
Multiwavelet Based
Compression
Compression
in %
Computation
Time
In Sec.
Mean
Square
Error
(MSE)
Compression
in %
Computation
Time
In Sec.
Mean
Square
Error
(MSE)
T-128 128x128
24 bits 35 23 0.18 67 07 0.1
T-256 256x256
24 bits 33 08 0.45 47 04 0.25
September Issue Page 68 of 105 ISSN 2229 5216
International Journal of Advances in Science and Technology,
Vol. 3, No.3, 2011
Figure 9. (A) Original image (B) Wavelet based output (C) multiwavelet based output
Figure.10. PSNR comparison plot
5. Conclusion
In this work it is observed that Hierarchical zero wavelet coding is able to achieve its
good performance with a relatively simple algorithm. Hierarchical zero wavelet coding does
not require complicated bit allocation procedures like subband coding does, it does not require
training or codebook storage like vector quantization does, and it does not require prior
knowledge of the image source like wavelet based compression does (to optimize quantization
tables). Hierarchical Zero Wavelet Coding also has the desirable property, resulting from its
successive approximation quantization.One desirable consequence of an embedded bit stream is
that it is very easy to generate coded outputs with the exact desired size. Truncation of the
September Issue Page 69 of 105 ISSN 2229 5216
International Journal of Advances in Science and Technology,
Vol. 3, No.3, 2011
coded output stream does not produce visual artifacts since the truncation only eliminates the
least significant refinement bits of coefficients rather than eliminating entire coefficients as is
done in subband coding. From the obtained results it is concluded that Hierarchal Zero wavelet
coding takes comparatively less(about 60%) time then the Wavelet Based Compression coding
system. The coding also shows less percentage of error in compressed image compare to the
existing Wavelet Based Compression coding system. It is observed that image coded with
Hierarchal Zero Wavelet tree coding shows clearer image than other coding system.
VI. REFERENCES
[1]. J.Shapiro, Embedded image coding using zerotrees of wavelet coefficients, IEEE Trans.
Signal Processing, vol. 41, pp.3445-3462, Dec 1993.
[2] S.Mallat and F. Falzon, Analysis of low bit rate image transform coding, IEEE Trans.
Signal Processing, vol. 46, pp. 1027-1042, Apr. 1998.
[3] Z. Xiong, K . Ramchandran, and M. Orchad, Space-frequency quantization for wavelet
image coding, IEEE Trans. Signal Processing, vol. 6, pp. 677-693, May 1997.
[4] E. H. Adelson, E. Simoncelli, and R. Hingorani, Orthogonal pyramid transforms for image
coding, Proc. SPIE, vol. 845, Cambridge, MA, Oct. 1987, pp. 50-58.
[5] R. A. DeVore, B. Jawerth, and B. J. Lucier, Image compression through wavelet transform
coding, IEEE Trans. Informat. Theory, vol. 38, pp. 719-746, Mar. 1992.
[6] S. Mallat, A theory for multiresolution signal decomposition: The wavelet representation,
IEEE Trans. Pattern Anal. Mach. Intell., vol. 37, pp.2091-2110, Dec.1990.
[7] G. K. Wallace, The JPEG Still Picture Compression Standard, Commun. ACM, vol. 34,
pp. 30-44, Apr. 1991.
[8] Bryan E. Usevitch, A Tutorial on Modern Lossy Wavelet Image Compression: Foundations of
JPEG 2000, IEEE signal processing magzine, 1053-5888,sep-2001.
Authors Profile
LalithaY.S was born on December 7, 1969 in India. She received B.E degree in
Electronics and Communication Engineering and M.E. degree in Power
Electronics from Gulbarga University Gulbarga, India, in 1991 and 2002
respectively. She is working as Principal, in Godutai Engineering college for
women, Gulbarga, India. Her research interests include image Processing, Wavelet
Transform coding. She attended Four National Conferences and three International
Conferences. Published 3 international journal papers.
Dr.Mrityunjaya V Latte was born on April 25
th
1964 in India. He received B.E.
degree in electrical Engineering and M.E. degree in Digital electronics from S.D.M. College of
Engineering & Technology, Dharwad, India in 1986 & 1995 respectively. He was awarded Ph.D.
degree in 2004 for his work in the area of digital signal processing. He is working as Principal, JSS
Academy of Technical Education, Bangalore, India. His research interests include coding, image
processing and multiresolution transforms. He received a best paper award for his paper in a national
conference NCSSS -2002 held at Coimbatore, India.
September Issue Page 70 of 105 ISSN 2229 5216