Time filter

Source Type

Hyderabad, India

Kranthi Kumar K.,SNIST | Gopal T.V.,JNTUH College of Engineering
2014 International Conference on Signal Propagation and Computer Technology, ICSPCT 2014

This paper, proposes a Non-Continuation based Self Re-Weighting approach for CBIR systems, to reduce semantic gap which is a bottle neck of CBIR. The assumption for previous FRW approaches are that the length of feature vectors for images are fixed and uses only the information from the set of images sent back in the early query result for feature re-weighting. The proposed system automatically recalculates the weight of features for an image, which estimate the user perception from the user feedback on retrieved set based on obtained interval. In this approach we examined systematically with other feature re-weighting methods and proved that our approach outperforms other approaches. Which we experimented with COREL database with 25 different categories and each category contains with 100 numbers of relevant images. The experimental results demonstrated the advantage of our approach in terms of precision and recall. © 2014 IEEE. Source

Kumar K. K.,SNIST | Gopal T.V.,JNTUH College of Engineering
Proceedings of 2014 International Conference on Contemporary Computing and Informatics, IC3I 2014

In this paper, we propose an approach using multilevel and multiple approaches for Feature Reweighting for CBIR system to reduce semantic gap using Relevance feedback. The first step of this approach does analysis on the positive and negative images, Second step calculates normalized feature component sets of images, Third step calculates overall distances between given query image and database images, and the next step calculates Relevance score along with confidence of the image, it is used for Feature Reweighting. All the above methods are performed individually in the previous systems, where as in our propose system we perform all these together. The assumption for the previous relevance feedback systems are that, all the above methods are performed against to the user given feedback. This increases the number of iterations for the retrieval systems. The propose system can do analysis of images, overall distance calculation, automatically calculates the weight of features for an image based on the confidence and score of the relevance before user feedback. And these results are carried forward to the next iteration for further calculations after the user feedback. © 2014 IEEE. Source

Jyothi B.V.,CBIT | Eswaran K.,SNIST
ISMS 2010 - UKSim/AMSS 1st International Conference on Intelligent Systems, Modelling and Simulation

In this paper, we describe two new approaches to content-based image retrieval (CBIR) based on preference information provided by the user interacting with an image search system. First, we present the existing methods of image retrieval with relevance feedback, which serve then as a reference for the new approaches. The first extension of the distance function-based CBIR approach makes it possible to apply this approach to complex objects .Next we discuss the second approach for image retrieval. That new algorithm is based on an approximation of user preferences by a neural network. Finally we discuss the advantages and disadvantages and further improvements and future scope in this particular area. © 2010 IEEE. Source

Ilamathi R.,SNIST | Nirmala G.S.,Vellore Institute of Technology | Muruganandam L.,Vellore Institute of Technology
International Journal of ChemTech Research

Our work aims to throw light on biosorption of heavy metals in a Liquid Solid Fluidized Bed as a successful alternative for heavy metal removal. The design and fabrication of LSFB has been discussed. Batch studies and fluidized bed studies were carried out to study the biosorption behavior for chromium, nickel, copper and cadmium by alginate beads containing a mixed consortium of Yeast, Pseudomonas aeruginosa, Bacillus subtilis and Escherichia coli. Fluidized bed studies were carried out in 1m length and 5cm diameter column, with an optimized adsorbent dosage of 1g/L, a flowrate of 132 LPH, a bed height of length of the reactor. Efficiency of biosorption for copper, cadmium, chromium and nickel in LSFB was found to be 84.62%, 67.17%, 49.25% and 61.02%. The efficiencies were found to depend on the pH, temperature, initial metal concentration, and the residence time of the beads in the fluidized beds. Desorption of the exhausted beads was successful, however, with a reduced biosorption capacity. Pretreatment of the culture was found to increase the capacity of metal uptake. Source

Bhutada S.,SNIST | Balaram V.V.S.S.S.,SNIST | Bulusu V.V.,JNTUHCEJ
International Journal of Applied Engineering Research

Dynamic topic extraction is a method helpful to understand the hidden knowledge from the textual database in order to systematize and supervise the growing text. The important challenge in this process is to insert new documents into appropriate categories. On the other hand misplacing of documents propagates the wrong information to the future topic hierarchy, thereby declining the quality of the knowledge extraction process. So, the effective extraction of topics and insertion of dynamic documents to a corresponding category is very important. Accordingly, a new method called, Dynamic Semantic Latent Dirichlet Allocation (DSLDA) is proposed in this paper by extending SLDA by handling the dynamic updates. Dynamic handling of documents requires much dimensional space to extract the feature words for dynamic process. In order to alleviate this problem, a method is developed using holoentropy which enables feature evaluation function to select the most important features from the document. The advantage of holoentropy is that, it can measure the global disorder of a data set by computing the total correlation to ensure the attribute relationship. These two proposals i.e. DSLDA and holoentropy are to be effectively integrated in the proposed system to dynamically handle the input documents and thereby, updating the topics using membership and representative information. The experimentation is performed using two different textual databases and the performance of the proposed DSLDA is validated using F-measure, Entropy, Rand and Jaccord Coefficient. © Research India Publications. Source

Discover hidden collaborations