Medical Image Processing - Digital Signal Processing for Medical Imaging Using Matlab - page 82

Fig. 4.10 Original image and the first level wavelet decomposed images using db 6filter

4.3.2 Using KL-Transformation

1. Consider the group of similar medical images to be compressed. Example: MRI
images.
2. Every column of the individual medical image is treated as the vector.
3. Compute the co-variance matrix C using the collected vectors.
4. Compute the significant eigen vectors of the C corresponding to the significant
eigen values.
5. Represent each column of the image to be compressed as the linear combinations
of significant vectors. The corresponding co-efficients are stored along with the
significant eigen vectors as the compressed data.
6. Significant eigen vectors are common for all the images. The corresponding co-
efficients along with the stored significant eigen vectors are used to reconstruct
the image back.
7. Number of significant eigen vectors determines the quality of the reconstructed
image.
8. 27 MRI images each with size 128 × 128 is considered to demonstrate the KLT

based compression.
9. Every image is reconstructed back with two different compression ratio and are
displayed in the Fig. 2.3 . One with 50 significant eigen vectors ( compression ratio
1:2.56) and another with 25 significant eigen vectors (compression ratio 1:5.12).