We present a wavelet-based algorithm to cut an object out of an image in a semi-automatic manner. The user of the algorithm selects a piece of a boundary that separates an object from the background. In contrast to other well-known segmentation algorithms, this sample boundary does not have to be a sharp edge, but might also be a smooth transition between two shades or a texture. Based on a multi-scale wavelet analysis of the sample area our algorithm seeks a path around the object by analyzing characteristic frequencies.
Dieser Eintrag ist Teil der Universitätsbibliographie.
Search Authors in