Information Technology Reference
In-Depth Information
in textual document retrieval systems. A new similarity distance function was adopted to measure the
similarity in the classification results. The query was performed by providing a query image from a
dataset that contains visual features (i.e., color) and five texture features (maximum probability [MP],
contrast [Cont], inverse difference moment [IM], angular second moment [AM], and entropy [Entro])
calculated from each occurrence matrix. Their values were saved in the feature vector of the correspond-
ing image. Then the rules were generated and ordered. The similarity between the images was estimated
by summing up the distance between the corresponding features in their feature vectors. Images having
feature vectors closest to the feature vector of the query image were returned as best matches. The results
were then numerically sorted, and the best images were displayed along with the query image.
All inputs to the rough set classifier (the image mixed textual and color vectors) were normalized,
and a set containing a minimal number of features (attributes) was contracted based on the rough set
criteria. Then we trained the model to have a mean of zero and variance of 1.
Our visual system then analyzed the sample images associated with each subtask. If the query im-
age was deemed to be a color image by the system, the set of top 50 textual images was processed and
those that were deemed to be color were moved to the top of the list. Within that list, the ranking was
based on the ranking of the textual results. The introduced image classification system was created using
MATLAB and ROSETTA software Figure 2 shows an example of the retrieved results.
Precision-recall measure is the conventional information retrieval performance measure. Precision
is defined as the number of relevant images retrieved relative to the total number of retrieved images,
Figure 2. Images retrieved
Search WWH ::




Custom Search