Abstract

The segmentation of objects coming from digital images is an area that is increasingly being addressed with greater emphasis. Diverse techniques have been applied to segment images obtained by Computer Tomography (CT). One of the principal problems is that the majority of these images are guided, principally by image intensity, creating complications when images possess noisy information. This work proposes an automatic segmentation method for images obtained by CT. It proposes the use of active contour with an energy function that weights the image's potential with object information known a priori to segmentation. To validate the proposed technique, wood pieces, saturated in water and noise, were inspected to isolate knots and other defects. The results have been surprising, obtaining very acceptable degrees of certainty, greater than those obtained with traditional techniques.