Abstract:Curves are important features in computer vision and pattern recognition, and their classification under a variety of transformations, such as Euclidean, affine or projective, poses a great challenge. Invariant features of these curves turn out to be crutial to simplifying any classification procedure. This, as a result, has recently led to a renewed research interest in transformation invariants. In this thesis, new explicit formulae for integral invariants for curves in 3D with respect to the special and the full affine groups are presented. The development of the 3D integral invariant are based on an inductive approach in terms of Euclidean invariants. For the first time, a clear geometric interpretation of both 2D and 3D integral invariants is presented. Since integration attenuates the effects of noise, integral invariants have advantages in computer vision applications. We use integral invariants to construct global and local signatures that characterize curves up to the special affine transformations, subsequently extended to the full affine group. Global Signatures are independent of parameterization, and Local Signatures are independent of both parameterization and initial point selection. We analyze the robustness of these invariants in their application to the problem of classification of noisy spatial curves extracted as characteristics from a 3D object. Our investigation of 2D and 3D integral invariants and signatures, originally motivated by Biometrics applications, are successfully implemented and applied to face recognition to eliminate the effects of pose and facial expression. A high recognition performance rate of 95% is achieved in the test with a large face data set.

Local Tangent Distances for Classification Problems

Image distance measurement based on manifold learning and manifold high-order approximation

Distance-Based Classifier Via the Kernel Trick.

Image Retrieval and Recognition Based on Generalized Local Distance Functions

Probabilistic Tangent Subspace

Tangent Space Learning and Generalization

Local Tangent Space Alignment Algorithm of Generalized Improvement

Improved Tangent Space Based Distance Metric for Accurate Lithographic Hotspot Classification

Saliency-Aware Image-To-Class Distances for Image Classification

Discriminative analysis algorithm for linear local tangent spaces

A local tangent space alignment based transductive classification algorithm

K-Ns: A Classifier by the Distance to the Nearest Subspace.

Combining intrinsic dimension and local tangent space for manifold spectral clustering image segmentation

Classification of Hyperspectral Image Based on Sparse Representation in Tangent Space

On the Translation-Invariance of Image Distance Metric

3D Mixed Invariant and Its Application on Object Classification

Supervised Local Tangent Space Alignment for Classification

Three-dimensional integral invariant signatures and their application on face recognition

Object Identification and Recognition Using Multiple Contours Based Moment Invariants

Tangent space spatial filters for interpretable and efficient Riemannian classification

Two-Dimensional Maximum Local Variation Based on Image Euclidean Distance for Face Recognition