On robust biometric identity verification via sparse encoding of faces: Holistic vs local approaches

Abstract

Abstract-In the field of face recognition, Sparse Representation (SR) has received considerable attention during the past few years. Most of the related literature focuses on holistic descriptors in closed-set identification applications. The underlying assumption in identification is that the gallery always has sufficient samples per subject to linearly reconstruct a query image. Unfortunately, such assumption is easily violated in the more challenging and realistic face verification scenario. A verification algorithm is required to determine if two faces (where one or both have not been seen before) belong to the same person, while explicitly taking into account the possibility of impostor attacks. In this paper, we first discuss why most of the SR literature is not applicable to verification problems. Motivated by the success of bag-of-words methods in the field of object recognition, which describe an image as a set of local patches or interest points, we then propose to tackle the verification problem by encoding each local face patch through SR. The locally encoded sparse vectors are pooled to form regional descriptors, where each descriptor covers a relatively large portion of the face. Experiments in various challenging conditions show that the proposed method achieves high and robust verification performance.

Action recognition using invariant features under unexampled viewing conditions

“…3) We showed that the SCAE can be efficiently applied to dense features such as positions and orientations for further refinement. In contrast to the sparseness constraints in sparse autoencoders [5,10], which expect the ratio of activated units in the hidden layer to be less than 0.1 and even close to 0, the reasonable range of our compact constraint ρ is between 0.4 and 0.7. This is because the input data in our case contain dense information, rather than the sparse information conveyed by image or video patches [2].…”

Automatic Classification of Human Epithelial Type 2 Cell Indirect Immunofluorescence Images using Cell Pyramid Matching

“…Nevertheless, the generalisation ability of these systems is still not guaranteed since these systems were only evaluated on a dataset with a specific setup. One of the most popular approaches for automatic image classification, here called the bag-of-visualwords (BoW) approach, is to represent an image in terms of a set of visual words, selected from a dictionary that has been trained beforehand [18,30,39,42]. In order to model an image, the BoW approach divides the image into small image patches, followed by patch-level feature extraction.…”

Effect of Edge Based Contours on Face Recognition

“…The ratio of determinants of these matrices is taken and the objective is to maximize this ratio. Holistic approaches primarily analyse the image as a whole instead of separating the background and foreground, the performance of the same is found in the literature [7] [8]. Although this approach is primitive, simple to implement and the least computationally intensive method of face recognition, it is also least tolerant to noise.…”

“…In order to account for the decreased performance of holistic approaches in cluttered scenes as reported by [7], several feature based approaches were developed. In this section we mainly discuss Gabor Based approach and Sparse Encoding technique.…”

“…Face recognition using Gabor feature [9][10] is performed by modulating a Gaussian kernel function in spatial domain by a sinusoidal plane wave. A more robust Sparse Encoding technique [7] for face recognition is performed by converting a full matrix of face image into sparse form by eliminating any non-zero entity and directing the resultant to zero. Motivated by the performance of feature based approaches [11] coupled with the effectiveness Sparse Encoding [7], we intuitively perceived that an image with only edge information would be more sparse, thereby increasing the PSR value [2] of the correlation filter output and enhancing the face recognition characteristics of the system.…”