Page 153 - Algorithm Collections for Digital Signal Processing Applications using MATLAB

142 Chapter 4

size 1x64. Represent this vector as the linear combination of Eigen
basis obtained as described in the step 4. The co-efficients are
obtained as the product of transformation matrix with the vector.
Transformation matrix is the matrix filled up with Eigen basis
arranged in the row wise. Number of Eigen co-efficients obtained is
equal to the dimension of the Ear vector space.

Step 8: Thus every 1x64 sized vector obtained from every sub blocks of the
ear image is mapped into the vector of size [1Xb], where ‘b’ is the
dimension of the ear vector space which is less than 64. Thus every
sub blocks of the ear image is stored with ‘b’ values which is very
much less than 64. In this experiment, the value of ‘b’ is
20<<64. Thus compression with the ratio 3.2:1 is achieved using
eigen basis technique.

Step 9: Every sub blocks of the decompressed image are obtained as the
linear combinations of Eigen basis with the corresponding co-
efficients associated with that sub block. The compressed and
decompressed image of the sample ear image is displayed in the
figure 4-3.

Figure 4-3. Ear image compression with the ratio 3.2:1 using Eigen basis