Feature extraction module

We can see feature extraction module on figure A.3 on page . The abstract module is the interface definition. The following must be implemented by ancestors:

• $\mathbf{\mu_I}$ = extract_features($I$)
• $\mathbf{\mu_n}$ = normalize( $\mathbf{\mu_I}$)

where $I$ is an input image, $\mathbf{\mu_I}$ is non-normalized feature vector and $\mathbf{\mu_n}$ is its normalized version. The normalize() function is useful since it makes a distribution from feature vector.

Figure A.3: Feature Extraction Module