22-05-2012, 10:52 AM
Shape Description for Content-based Image Retrieval
[attachment=22659]
Introduction
Images indexing and retrieval using content has gained increasing importance
during last years. Almost all kinds of image analysis techniques have been investigated
in order to derive sets of meaningful features which could be useful
for the description of pictorial information, and a considerable effort has been
spent towards the development of powerful but easy-to-use commercial database
engines.
The most popular CBIR (content-based image retrieval) systems developed
so far, like QBIC [5, 7] Photobook [9], Virage [8], model the image content as
a set of uncorrelated shape, texture and color features. Queries are obtained
either by manual specification of the weights for each feature or by presenting
an example to the system, and they’re refined by means of various relevance
feedback strategies.
Theoretical Remarks
The use of KLT features for pattern recognition is a well known technique in
the computer vision community [10, 12] but, in general, this transformation is
applied to the image as a whole, and the transformed vectors are used directly
for the classification task.
In our approach, KLT is implicitly applied to the scan-lines of a generic subimage,
and only the eigenvalues of their covariance matrix are taken into account.
In other words, given a N ×M rectangular region of an image, we compute the
matrix:
Description of the System
In this section, the complete structure of the presented system will be reported,
starting from the considerations of the preceding section.
We’ve analyzed the histogram of the components of λ computed from several
images both synthetic and real, depicting single shapes under varying attitudes
and lighting (see figure 1). We’ve noticed that this histogram exhibits some
dominant modes, whose relative position and amplitude depend on the shape
observed. The amplitude and position of these histogram modes remain almost
unchanged under rotation, translation, and scaling of the object.
Automatic Search Algorithm
The search algorithm we implemented is based on a two-pass strategy. The first
step performs rough location of the ROIs both for the horizontal and vertical displacement.
The second step defines the windows’ dimensions for all the selected
positions.
[attachment=22659]
Introduction
Images indexing and retrieval using content has gained increasing importance
during last years. Almost all kinds of image analysis techniques have been investigated
in order to derive sets of meaningful features which could be useful
for the description of pictorial information, and a considerable effort has been
spent towards the development of powerful but easy-to-use commercial database
engines.
The most popular CBIR (content-based image retrieval) systems developed
so far, like QBIC [5, 7] Photobook [9], Virage [8], model the image content as
a set of uncorrelated shape, texture and color features. Queries are obtained
either by manual specification of the weights for each feature or by presenting
an example to the system, and they’re refined by means of various relevance
feedback strategies.
Theoretical Remarks
The use of KLT features for pattern recognition is a well known technique in
the computer vision community [10, 12] but, in general, this transformation is
applied to the image as a whole, and the transformed vectors are used directly
for the classification task.
In our approach, KLT is implicitly applied to the scan-lines of a generic subimage,
and only the eigenvalues of their covariance matrix are taken into account.
In other words, given a N ×M rectangular region of an image, we compute the
matrix:
Description of the System
In this section, the complete structure of the presented system will be reported,
starting from the considerations of the preceding section.
We’ve analyzed the histogram of the components of λ computed from several
images both synthetic and real, depicting single shapes under varying attitudes
and lighting (see figure 1). We’ve noticed that this histogram exhibits some
dominant modes, whose relative position and amplitude depend on the shape
observed. The amplitude and position of these histogram modes remain almost
unchanged under rotation, translation, and scaling of the object.
Automatic Search Algorithm
The search algorithm we implemented is based on a two-pass strategy. The first
step performs rough location of the ROIs both for the horizontal and vertical displacement.
The second step defines the windows’ dimensions for all the selected
positions.