IJSRST162663 | Received: 01 Dec 2016 | Accepted: 08 Dec 2016 | November-December-2016 [(2)6: 340-342 ]
© 2016 IJSRST | Volume 2 | Issue 6 | Print ISSN: 2395-6011 | Online ISSN: 2395-602X
Themed Section: Engineering and Technology
340
Entropy Optimization for Image Retrieval
Sona Rahim, Hafsath Ca
Department of Computer Science & Engineering, Ilahia College of Engineering & Technology, Muvattupuzha, Kerala, India
ABSTRACT
Recent studies shows that there are several information retrieval are being developed. Earlier the main focus was
text retrieval. Information retrieval (IR) concerned with the searching and retrieving of knowledge-based
information from database. In this paper, we represent the various models and techniques for information retrieval.
This paper focuses some of the most related image retrieval techniques.
Keywords : Search And Retrieval, Dictionary Learning, BOW, Entropy Optimization, Image Retrieval.
I. INTRODUCTION
Information retrieval (IR) is the task of retrieving objects,
like images, from a database where the user's
information need. Research focused mostly on text
retrieval ,now in image retrieval, and video retrieval ,
since the availability of digital technology led to a great
increase of multimedia data Image retrieval is the most
studied and challenging aspect of multimedia
information retrieval.
For image category classification Bag of words process
is used,it is by extracting feature of images.Bag of words
is a model that is used in natural language processing
and information retrieval.The BoW model takes each
image as a document ,it contains a number of different
“visual” words.The features can be extracted by some
clustering algorithms.From this bag of words(BOW)
codebook is generated
In BOW features are calculated in terms of
frequency.Most popular ways to normalize the
frequencies is to weight a term by inverse of document
frequency(tf-idf).For classification purpose class label of
a document is taken.Entropy is a key measure in
Information retrieval,it quantifies the amount of
uncertainty in values or the outcome of random
process.Entropy refers to the disorder . The state-of-the
art methods, b) reduce the storage requirements and
query time by using smaller dictionaries, and c) transfer
the learned knowledge to previously unseen classes
without retraining.
The EO-BoW, which optimizes a retrieval-oriented
objective function. We demonstrated the ability of the
proposed method to improve the retrieval performance
using two image datasets, a collection of time-series
datasets, a text dataset and a video dataset.
II. METHODS AND MATERIAL
Entropy optimization by bag of words, in this method
bag of words alone cannot be used, it can be used for
extracting a feature vector from each word of a text
document .For optimizing the Entropy is opted ,inoder to
maximize the relavant information[1].It is a effective
retrieval dictionary learning method to improve retrieval
precision beyond other methods and can reduce storage
space and time .Without training learned dictionary can
be transfered.By using entropy codebook can be
optimized.By this method images can be retrieved and
can be used in other retrieval.
Spatial Pyramid Matching for categorization,in this
paper the semantic category of the image is resolved by
spatial pyramid matching method.It a method for
effective scene categorization[2].It produce high
accuracy on large database. Build pyramid in image
space, quantize feature space is the approach used. Find
maximum-weight matching (weight is inversely
proportional to distance). Global spatial regularities
(natural scene statistics) help even in databases with
high geometric variability.