On Utilising Template and Feature-Based Correspondence in Multi-view Appearance Models

摘要: In principle, the recovery and reconstruction of a 3D object from its 2D view projections require parameterisation shape structure surface reflectance properties. Explicit representation such information is notoriously difficult to achieve. Alternatively, linear combination views can be used which requires establishment dense correspondence between views. This in general, compute necessarily expensive. this paper we examine use affine local feature-based transformations establishing correspondences very large pose variations. doing so, utilise generic-view template, generic model Kernel PCA for modelling texture nonlinearities across The abilities both approaches reconstruct recover faces any image are evaluated compared.