Neural Computation
The nature of statistical learning theory
The nature of statistical learning theory
Computable elastic distances between shapes
SIAM Journal on Applied Mathematics
Natural gradient works efficiently in learning
Neural Computation
Diffeomorphisms Groups and Pattern Matching in Image Analysis
International Journal of Computer Vision
Nonlinear component analysis as a kernel eigenvalue problem
Neural Computation
An introduction to support Vector Machines: and other kernel-based learning methods
An introduction to support Vector Machines: and other kernel-based learning methods
Content-Based Image Retrieval at the End of the Early Years
IEEE Transactions on Pattern Analysis and Machine Intelligence
Geometry-Driven Diffusion in Computer Vision
Geometry-Driven Diffusion in Computer Vision
Choosing Multiple Parameters for Support Vector Machines
Machine Learning
Modal Matching for Correspondence and Recognition
IEEE Transactions on Pattern Analysis and Machine Intelligence
IEEE Transactions on Image Processing
| Hi-index | 0.00 |
The design of good features and good similarity measures between features plays a central role in any retrieval system. The use of metric similarities (i.e. coming from a real distance) is also very important to allow fast retrieval on large databases. Moreover, these similarity functions should be flexible enough to be tuned to fit users behaviour. These two constraints, flexibility and metricity are generally difficult to fulfill. Our contribution is two folds: We show that the kernel approach introduced by Vapnik, can be used to generate metric similarities, especially for the difficult case of planar shapes (invariant to rotation and scaling). Moreover, we show that much more flexibility can be added by non-rigid deformation of the induced feature space. Defining an adequate Bayesian users model, we describe an estimation procedure based on the maximisation of the underlying log-likehood function.