A short term and long term learning based on Fuzzy Transaction Repository and feature re-weighting

Abstract

In this paper, we introduce a combined relevance feedback approach for image retrieval using semantic similarity based on fuzzy transaction repository and feature re-weighting technique. This system accumulates user interactions using soft feedback model to construct fuzzy transaction repository (FTR). The repository remembers the userpsilas intent and therefore provides a better representation of each image in the database in terms of the semantic meanings. The semantic similarity between the query image and each database image can then be computed using the current feedbacks and the semantic values in the FTR. Furthermore, feature re-weighting is applied on the session-term feedback to learn weight of low level features. Then we use the weighted Euclidean distance metric to measure the distance between the query image and each database image. These two similarity measures are normalized and combined together to form the overall similarity measure. Our experimental results show that the average precision of the proposed system exceeds 83% after three iterations.