The model-based segmentation technique discussed in Section 16.5 is a first step toward integrating global information into the process of object recognition. It is an inflexible technique, however, as it requires an exact parameterization of the objects to be detected. For real objects, it is often not possible to establish such an explicit type of model.
KeywordsAnisotropy Covariance Resid Convolution Pyramid
Unable to display preview. Download preview PDF.