A Relevance Feedback Algorithm Combining Bayesian and FSRM

ABSTRACT

The semantic gap between low level visual features and high level semantic concepts, is an obstacle to the development of image retrieval. The semantic gap is narrowed by relevant feedback techniques to some extent. However, the image retrieval process with the relevant feedback technology also has many disadvantages such as too many feedback times or unsatisfactory feedback effect. In order to improve the relevance feedback method, a new relevance feedback strategy combining Bayesian and FSRM technology has been presented. The main approach was achieved firstly by assorting the image library with the Bayesian classifier compressing the image library; secondly, by searching the compressed image library with the FSRM; and lastly, by returning the worked out results. The experiment results illustrated the accuracy of the feedback method and showed it to be the best compared with FSRM algorithm and Bayesian algorithm.

Keywords:

Bayesian(method), fuzzy semantic relevance matrix(FSRM), image retrieval, normal distribution, relevance feedback.