Feature Matching Theory

1553 Words7 Pages

Numerous theories concerning how we identify and recognise objects are debated today. Template matching, feature matching, and structural analysis, all theories of object recognition, suggest how our brains interpret sensory input through the visual cortex and connect this input to meaning. To recognise an object, the brain matches the information passed through the retina to existing knowledge stored in our memory.
Perhaps, when presented with an object, the brain flicks through stored image templates (Tarr & Vuong, 2002) that ‘represent known objects’ (Edelman & Bulthoff, 1992, p. 2385) until a match to the stimulus is found, causing you to recognise the object. The name and associations of each object’s template may also be stored in the …show more content…

Rather than holistic template analysis, visual features are analysed separately, matched to known feature templates, and then to template combinations that occur in different objects. The brain then matches this combination to pre-existing knowledge of an object. One may argue the brain does not have the capacity to store all the separate components of every single object. However, the need for only a small number of components to identify objects combats this; Riesenhuber and Poggio (2000) and research by Gibson (1969) suggest requirements for features to be most effective in object recognition, including the need for the features to easily differentiate from one another so elimination of other potential combinations is easier, however, ensuring the number of components stored in the brain remains few, another requirement shows the need for basic templates, usable for many objects. This should provide a small of a range of features, still allowing efficient and effective identification (Gibson, 1969). In addition, some ideas involving structural theory suggest normalisation occurs after object deconstruction (e.g Edelman & Bulthoff, 1992; Ullman, 1989); the object components are readjusted in the brain to a mimic familiar viewpoint so identification can take

Open Document