Abstract: In this paper three different approaches for person
verification and identification, i.e. by means of fingerprints, face and
voice recognition, are studied. Face recognition uses parts-based
representation methods and a manifold learning approach. The
assessment criterion is recognition accuracy. The techniques under
investigation are: a) Local Non-negative Matrix Factorization
(LNMF); b) Independent Components Analysis (ICA); c) NMF with
sparse constraints (NMFsc); d) Locality Preserving Projections
(Laplacianfaces). Fingerprint detection was approached by classical
minutiae (small graphical patterns) matching through image
segmentation by using a structural approach and a neural network as
decision block. As to voice / speaker recognition, melodic cepstral
and delta delta mel cepstral analysis were used as main methods, in
order to construct a supervised speaker-dependent voice recognition
system. The final decision (e.g. “accept-reject" for a verification
task) is taken by using a majority voting technique applied to the
three biometrics. The preliminary results, obtained for medium
databases of fingerprints, faces and voice recordings, indicate the
feasibility of our study and an overall recognition precision (about
92%) permitting the utilization of our system for a future complex
biometric card.
Abstract: For over a decade, the Pulse Coupled Neural Network
(PCNN) based algorithms have been successfully used in image
interpretation applications including image segmentation. There are
several versions of the PCNN based image segmentation methods,
and the segmentation accuracy of all of them is very sensitive to the
values of the network parameters. Most methods treat PCNN
parameters like linking coefficient and primary firing threshold as
global parameters, and determine them by trial-and-error. The
automatic determination of appropriate values for linking coefficient,
and primary firing threshold is a challenging problem and deserves
further research. This paper presents a method for obtaining global as
well as local values for the linking coefficient and the primary firing
threshold for neurons directly from the image statistics. Extensive
simulation results show that the proposed approach achieves
excellent segmentation accuracy comparable to the best accuracy
obtainable by trial-and-error for a variety of images.
Abstract: In this paper, we propose a novel approach for image
segmentation via fuzzification of Rènyi Entropy of Generalized
Distributions (REGD). The fuzzy REGD is used to precisely measure
the structural information of image and to locate the optimal
threshold desired by segmentation. The proposed approach draws
upon the postulation that the optimal threshold concurs with
maximum information content of the distribution. The contributions
in the paper are as follow: Initially, the fuzzy REGD as a measure of
the spatial structure of image is introduced. Then, we propose an
efficient entropic segmentation approach using fuzzy REGD.
However the proposed approach belongs to entropic segmentation
approaches (i.e. these approaches are commonly applied to grayscale
images), it is adapted to be viable for segmenting color images.
Lastly, diverse experiments on real images that show the superior
performance of the proposed method are carried out.
Abstract: In current common research reports, salient regions
are usually defined as those regions that could present the main
meaningful or semantic contents. However, there are no uniform
saliency metrics that could describe the saliency of implicit image
regions. Most common metrics take those regions as salient regions,
which have many abrupt changes or some unpredictable
characteristics. But, this metric will fail to detect those salient useful
regions with flat textures. In fact, according to human semantic
perceptions, color and texture distinctions are the main characteristics
that could distinct different regions. Thus, we present a novel saliency
metric coupled with color and texture features, and its corresponding
salient region extraction methods. In order to evaluate the
corresponding saliency values of implicit regions in one image, three
main colors and multi-resolution Gabor features are respectively used
for color and texture features. For each region, its saliency value is
actually to evaluate the total sum of its Euclidean distances for other
regions in the color and texture spaces. A special synthesized image
and several practical images with main salient regions are used to
evaluate the performance of the proposed saliency metric and other
several common metrics, i.e., scale saliency, wavelet transform
modulus maxima point density, and important index based metrics.
Experiment results verified that the proposed saliency metric could
achieve more robust performance than those common saliency
metrics.
Abstract: In this work a novel approach for color image
segmentation using higher order entropy as a textural feature for
determination of thresholds over a two dimensional image histogram
is discussed. A similar approach is applied to achieve multi-level
thresholding in both grayscale and color images. The paper discusses
two methods of color image segmentation using RGB space as the
standard processing space. The threshold for segmentation is decided
by the maximization of conditional entropy in the two dimensional
histogram of the color image separated into three grayscale images of
R, G and B. The features are first developed independently for the
three ( R, G, B ) spaces, and combined to get different color
component segmentation. By considering local maxima instead of the
maximum of conditional entropy yields multiple thresholds for the
same image which forms the basis for multilevel thresholding.
Abstract: This paper is concerned with an improved algorithm
based on the piecewise-smooth Mumford and Shah (MS) functional
for an efficient and reliable segmentation. In order to speed up
convergence, an additional force, at each time step, is introduced
further to drive the evolution of the curves instead of only driven by
the extensions of the complementary functions u + and u - . In our
scheme, furthermore, the piecewise-constant MS functional is
integrated to generate the extra force based on a temporary image that
is dynamically created by computing the union of u + and u - during
segmenting. Therefore, some drawbacks of the original algorithm,
such as smaller objects generated by noise and local minimal problem
also are eliminated or improved. The resulting algorithm has been
implemented in Matlab and Visual Cµ, and demonstrated efficiently
by several cases.
Abstract: Deformable active contours are widely used in
computer vision and image processing applications for image
segmentation, especially in biomedical image analysis. The active
contour or “snake" deforms towards a target object by controlling the
internal, image and constraint forces. However, if the contour
initialized with a lesser number of control points, there is a high
probability of surpassing the sharp corners of the object during
deformation of the contour. In this paper, a new technique is
proposed to construct the initial contour by incorporating prior
knowledge of significant corners of the object detected using the
Harris operator. This new reconstructed contour begins to deform, by
attracting the snake towards the targeted object, without missing the
corners. Experimental results with several synthetic images show the
ability of the new technique to deal with sharp corners with a high
accuracy than traditional methods.