Medical Image Processing - Digital Signal Processing for Medical Imaging Using Matlab - page 84

Table 4.1 Discrete Wavelet
Transformation (DWT)

Row-wise Column-wise Type1

Low pass Low pass Approximation

Low pass High pass Horizontal

High pass Low pass Vertical

High pass High pass Diagonal

4.4.1 Using Discrete Wavelet Transformation

1. Discrete wavelet transformation (DWT) consists of four discrete 1 D -filters. They
are low pass decomposition filter, high pass decomposition, corresponding low
pass reconstruction filter and high pass reconstruction filter.
2. Every row of the medical image under consideration is passed through the 1 D -
filter 1.
3. Every column of the obtained image is passed through the 1 D -filter 2.
4. The alternative samples are removed from the obtained image to obtain the
Wavelet decomposed image.
5. Based on the combinations of the filter-1 and filter-2 selected from the low-pass
and the high-pass decomposition filters, we obtain approximation, horizontal,
vertical and the diagonal images of first level decomposition (refer Table 4.1 ).
6. The approximation of the first level decomposed image is treated as the original
image and are subjected to decomposition to obtain second level decomposition
(refer Fig. 4.11 ) and so on.
7. The decomposed values (significant) of last level approximation image and all the
horizontal, vertical and diagonal values (upto to the last level decomposition) are
collected as the vector and are treated as feature vector for further classification.
8. Based on the type of 1 D filters used, we obtain different wavelet decomposition.
The db 61 D -filters are used for illustration.
9. Corresponding recomposition filters are used for DWT basedmedical image com-
pression and denoising. Decomposition filters are not used for feature extraction.

4.4.1.1 feaextwavelet.m

load mri
A=D(:,:,1,14);
[C,L]=wavedec2(A,2,'db6');
A=appcoef2(C,L,'db6',2);
[H2,V2,D2]=detcoef2('all',C,L,2);
[H1,V1,D1]=detcoef2('all',C,L,1);

figure
colormap(gray)