A Robust PCA Algorithm for Building Representations from Panoramic Images

Abstract

Abstract. Appearance-based modeling of objects and scenes using PCA has been successfully applied in many recognition tasks. Robust methods which have made the recognition stage less susceptible to outliers, occlusions, and varying illumination have further enlarged the domain of applicability. However, much less research has been done in achieving robustness in the learning stage. In this paper, we propose a novel robust PCA method for obtaining a consistent subspace representation in the presence of outlying pixels in the training images. The method is based on the EM algorithm for estimation of principal subspaces in the presence of missing data. By treating the outlying points as missing pixels, we arrive at a robust PCA representation. We demonstrate experimentally that the proposed method is efficient. In addition, we apply the method to a set of panoramic images to build a representation that enables surveillance and view-based mobile robot localization.

Generic visual categorization using composite Gabor and moment features

“…Therefore, objects can be described into different types. These include model-based approaches [2][3][4], shape-based approaches [5,6] and appearancebased approaches [7][8][9][10]. Model-based approaches try to represent the object as a collection of three dimensional, geometrical primitives (boxes, spheres, cones, cylinders, generalized cylinders, surface of revolution) whereas shape-based methods represent an object by its shape/contour.…”

Fast-Robust PCA

“…Due to its least squares formulation, PCA is highly sensitive to outliers. Thus, several methods for robustly learning PCA subspaces (e.g., [12,13,14,15,16]) as well as for robustly estimating the PCA coefficients (e.g., [17,18,19,20]) have been proposed. In this paper, we are focusing on the latter case.…”

Face Recognition with Weighted Locally Linear Embedding

“…Skočaj et al [12] propose a technique to extend PCA to include weights for images in the training set as well as individual pixels in the image. These weights are used both in the recognition of the test face, and the computation of the reduced space.…”

“…Skočaj et al [12] have applied weighted PCA to image recognition, using two types of weights: weights for individual pixels (spatial weights), used to account for parts of the image which are unreliable or unimportant, and weights for images, which they refer to as temporal weights. The main motivation for the latter is the idea that more recent images of a subject (or object) are more reliable.…”