Template matching approach to content based image indexing by low dimensional Euclidean embedding
Haim Schweitzer · 2002
Content based indexing is computed from input that consists of matching values between images and templates. The key idea is to embed both images and templates in a low-dimensional Euclidean space so that matching between embedded images and embedded templates approximates the given input. It is shown that such embedding can be computed by means of a singular value decomposition of the input matrix. Classic principal component analysis is shown to be a special case of the proposed technique, corresponding to the case where the templates and the images are the same.