Re-Ranking Of Web Images Using Semantic Signature and Parallel SVM
Journal Title: International Journal for Research in Applied Science and Engineering Technology (IJRASET) - Year 2015, Vol 3, Issue 6
Abstract
Image re-ranking has been adopted by current commercial search engines to improve the results of web-based image search. In image re-ranking, for a given a query keyword, first a pool of images are retrieved by the search engine and the user is asked to select a query image from the pool. The remaining images are re-ranked based on their visual similarities with the query image. But there are difficulties in understanding users’ search intention and in learning a universal visual semantic space to characterize the web images. An image re-ranking framework is proposed which automatically offline learns different visual semantic spaces for different query keywords through keyword expansions. The visual features of images are projected into their related visual semantic spaces to get semantic signatures. At the online stage, images are re-ranked by comparing their semantic signatures obtained from the visual semantic space specified by the query keyword. An image can have a single semantic signature or multiple semantic signatures for a specific keyword. Computing multiple semantic signatures for an image has higher precision and re-ranking accuracy than using single semantic signature. But the lengthy combined semantic signatures results in more time consumption. Hence a number of SVMs (Support Vector Machines) can be used in parallel at the online stage which compares multiple semantic signatures at the same time and results are merged at the later stages. Thus images are re-ranked in faster manner while preserving the accuracy
Authors and Affiliations
Nisha Abdul, Kiran V K
Keywords
Related Articles
- EP ID EP21028
- DOI -
- Views 292
- Downloads 6
Cholesterol Content and Free Fatty Acids in Edible Oils and Health Effects: A review
Cholesterol plays a major role in human heart health and high cholesterol is a leading risk factor for human cardiovascular diseases such as coronary heart disease and stroke. Not only cholesterol content but also chole...
Mining of infrequent itemset From Transactional Weighted Datasets Using Frequent Pattern Growth
Itemset mining is a data mining method extensively used for learning important correlations among data. Initially item sets mining was made on discovering frequent itemsets. Frequent weighted item set characterizes data...
Implementation of K-Means Clustering Algorithm in Hadoop Framework
Drastic growth of digital data is an emerging area of concern which has led to concentration of Data Mining technique. The actual data mining task involves programmatic or semi-programmatic analysis of large quantities...
slugInvestigation of Thermal Boundary Layer of non-Newtonian Fluid Past over a Wedge
This paper investigates the problem of steady, laminar, two-dimensional boundary layer flow and heat transfer of an incompressible, viscous non-Newtonian fluid over a wedge. A non-Newtonian fluid represented by power la...
Comparison of Effectiveness of Two Different Setups of Double Pipe Heat Exchangers
At present cost and performance are the two major factors which plays a vital role in the selection of any augmentation method of heat transfer. Cutting of triangular shaped fins is also very difficult to insert inside...