FLEXIBLE IMAGES - MATCHING AND RECOGNITION USING LEARNED DEFORMATIONS

We describe a novel technique for matching and recognition based on de formable intensity surfaces which incorporates both the shape (x, y) a nd the texture (I(x, y)) components of a 2D image. Specifically, the i ntensity surface is modeled as a deformable 3D mesh in (x, y, I(x, y)) space which obeys Lagrangian dynamics. Using an efficient technique f or matching two surfaces (in terms of the analytic modes of vibration) , we can obtain a dense correspondence field (or 3D warp) between two images, Furthermore, we use explicit statistical learning of the class of valid deformations in order to provide a priori knowledge about ob ject-specific deformations, The resulting formulation leads to a compa ct representation based on the physically-based modes of deformation a s well as the statistical modes of variation observed in actual traini ng data. We demonstrate the power of this approach with experiments ut ilizing image matching, interpolation of missing data, and image retri eval in a large face database. (C) 1997 Academic Press.