Pit Pattern Classification of Zoom-Endoscopical Colon Images Using DCT and FFT

This work presents a classification approach for images taken from magnifying colonoscopy. Classification is done according to the pit pattern scheme. Images are not classified directly in the proposed classifier. Instead, they are transformed to a frequency domain using discrete cosine or Fourier transformation. Feature selection is optimized using a genetic algorithm, the actual classification is done using standard methods from statistical pattern recognition (a Bayes normal classifier).