Skip to Main Content
One of rudimentary problems in Content-Based Image Retrieval (CBIR) has been the gap between low-level visual features and high-level semantic concepts. Relevance feedback (RF) is used to reduced this gap. In this paper Gausssian Mixture Model(GMM) is used to model the target distribution of query. Here a novel idea to estimate the components of GMM is proposed based on Connected component analysis. Connected component analysis uses positive and negative labeled examples obtained from relevance feedback to estimate the number of components of GMM. The retrieval performance of the proposed method is compared with MARS, and Mind Reader to show the efficiency using Wang and 5000 corel database.