PROJECT TITLE :

A-Optimal Projection for Image Representation

ABSTRACT:

We tend to consider the matter of image illustration from the angle of statistical style. Recent studies have shown that pictures are probably sampled from a coffee dimensional manifold despite of the very fact that the ambient space is usually terribly high dimensional. Learning low dimensional image representations is crucial for many image processing tasks such as recognition and retrieval. Most of the prevailing approaches for learning low dimensional representations, like principal component analysis (PCA) and locality preserving projections (LPP), aim at discovering the geometrical or discriminant structures in the data. During this paper, we take a different perspective from statistical experimental design, and propose a novel dimensionality reduction algorithm known as A-Optimal Projection (AOP). AOP is predicated on a linear regression model. Specifically, AOP finds the optimal basis functions so that the expected prediction error of the regression model can be minimized if the new representations are used for training the model. Experimental results suggest that the proposed approach provides a better representation and achieves higher accuracy in image retrieval.