SHAPE-DESCRIPTION AND RECOGNITION USING THE HIGH-ORDER MORPHOLOGICAL PATTERN SPECTRUM

Shape analysis and recognition is an important issue in modern image p rocessing and computer vision. It needs to solve three problems: possi bility, sensitivity and robustness. To deal with the third one, this p aper proposes the high order morphological pattern (HP) spectrum which relies upon the difference between two different sized alternating se quential morphological transformations on the original image. The HP s pectrum is characterized by the translating invariance and rotating in variance for the isotropic; structuring element. It is also proved tha t the higher order components of HP spectrum for a noisy image are exa ctly identical to these for the noise-free image under certain conditi ons. Also, the results of the experiments conducted, have shown that t he HP spectrum shows the strongest robustness in a noisy environment.