To look for examples of a particular class of objects in an image we can use a parameterized model. This might represent the relative positions of edge segments belonging to the object, or might describe the intensity variation over a region. To use the model to locate an object in an image we define an objective function, preferably a likelihood type function, which will allow us to estimate the quality of fit of a model instance to the image data given a set of parameters. Usually the model will have its own intrinsic parameters, and there will be additional extrinsic parameters describing the position, scale, orientation etc of the model in the image. We must choose the optimal set of intrinsic and extrinsic parameters to locate an object.