07-07-2012, 09:56 AM
ROBUST FACE RECOGNITION USING LOCALLY ADAPTIVE SPARSE
REPRESENTATION
ROBUST FACE RECOGNITION.pdf (Size: 172.99 KB / Downloads: 92)
ABSTRACT
This paper presents a block-based face-recognition algorithm based
on a sparse linear-regression subspace model via locally adaptive
dictionary constructed from past observable data (training samples).
The local features of the algorithm provide an immediate
benefit – the increase in robustness level to various registration
errors. Our proposed approach is inspired by the way human beings
often compare faces when presented with a tough decision:
we analyze a series of local discriminative features (do the eyes
match? how about the nose? what about the chin?...) and then
make the final classification decision based on the fusion of local
recognition results. In other words, our algorithm attempts to
represent a block in an incoming test image as a linear combination
of only a few atoms in a dictionary consisting of neighboring
blocks in the same region across all training samples.
INTRODUCTION
Sparse representations have been recently exploited in many pattern
recognition applications [1–3]. These approaches are based
on the assumption that a test sample approximately lies in a lowdimensional
subspace spanned by the training data and thus can be
compactly represented by a few training samples. The recovered
sparse vector then can be used directly for recognition. This approach
is simple and fast since no training stage is needed and the
dictionary can be easily expanded by additional training samples.
The original sparsity-based face recognition algorithm [1] yields
superior recognition performance comparing to existing techniques.
However, it suffers from the limitation that the test face must be
perfectly aligned to the training data prior to classification. To
overcome this problem, various methods have been proposed for
simultaneously optimizing the registration parameters and the sparse
coefficients [4, 5], leading to even more complicated systems.
SIMULATION RESULTS
In this section, we apply the proposed block-based algorithm for
identification on a publicly available database - the Extended Yale
B Database [11], and compare the performance with the original
algorithm in [1]. This database consists of 2414 perfectly-aligned
frontal face images of size 192 ×168 of 38 individuals, 64 images
per individual, under various conditions of illumination. In
our experiments, for each subject we randomly choose 15 images
in Subsets 1 and 2, which were taken under less extreme lighting
conditions, as the training data. Then, we randomly choose 500
images from the remaining images as test data. All training and
test samples are downsampled to size 32×28. The Subspace Pursuit
algorithm [8] is used to solve the sparse recovery problem (4).
CONCLUSION
In this paper, we propose a block-based algorithm for face recognition
via sparse representation. By constructing locally adaptive
dictionaries that capture the relative stationary features in a small
neighborhood, the proposed algorithm is robust to various types
of misalignment between the test and training data, without explicit
computation of the registration parameters. We propose to
use multiple blocks in the same test image and combine all classification
results to further improve the robustness. As demonstrated
by the simulation results on the Extended Yale B Database, the
proposed algorithm yields excellent performance in the presence
of registration errors.