Volume : II, Issue : X, October - 2013

Image Retrieval By Semi Supervised Biased Maximum Margin Analysis

Y. Anjinaiah, D. K. Jawad , D. Venkatesh

Abstract :

Many real time applications need image search. Such applications facilitate Query by Example (QBE). User gives an input image and the system retrieves relevant images. The features of input image are used in the query processing. This concept is known as CBIR (Content Based Image Retrieval). Considerable research has been carried out in this area to improve the accuracy of image retrieval. The existing systems used relevant feedback to improve the quality of CBIR. Many RF based schemes came into existence. Out of them SVM based RF has become popular. Its drawback is that it does not see difference between positive and negative feedbacks. Moreover it also ignores unlabeled samples. The SemiBMMA (Semi Supervised Biased Maximum Margin Analysis) proposed by Shang et al. addresses this problem. This paper further improves SemiBMMA by integrating it with Navigation Pattern Based Relevance Feedback (NPRF). Experiments made with prototype application revealed that the SemiBMMA with NPRF requires very less number of iterations to get user–satisfied results.

Keywords :

Article: Download PDF   DOI : 10.36106/ijsr  

Cite This Article:

Y. ANJINAIAH, D.K. JAWAD, D.VENKATESH / Image Retrieval By Semi Supervised Biased Maximum Margin Analysis / International Journal of Scientific Research, Vol.2, Issue.10 October 2013

Number of Downloads : 582

References :