Averaged Template Matching Equations

By exploiting an analogy with averaging procedures in fluid dynamics, we present a set of averaged template matching equations. These equations are analogs of the exact template matching equations that retain all the geometric properties associated with the diffeomorphism group, and which are expected to average out small scale features and so should, as in hydrodynamics, be more computationally efficient for resolving the larger scale features. From a geometric point of view, the new equations may be viewed as coming from a change in norm that is used to measure the distance between images. The results in this paper represent first steps in a longer term program: what is here is only for binary images and an algorithm for numerical computation is not yet operational. Some suggestions for further steps to develop the results given in this paper are suggested.

• Type: Conference
• Year: 2001
• Where: EMMCVPR
• Authors: Anil N. Hirani, Jerrold E. Marsden, James Arvo
• Keywords: Averaged Template, Computer Vision, EMMCVPR 2001, Scale Features, Template Matching Equations