Learning of Relevance Feedback Using a Novel Kernel Based Neural Network

In this paper, we introduce a novel neural network and fuzzy transaction based image retrieval system. The proposed system is a composite relevance feedback approach for image retrieval using semantic and visual learning. In semantic learning, the system integrates the log information of user feedback using a fuzzy feedback model to construct fuzzy repository. The repository remembers the userʹs intent and therefore, provides a better representation of each image in the database. The semantic similarity between the query image and each database image can then be computed using the current feedback and the semantic values in the fuzzy repository. In addition, most of the existing approaches assume a linear relationship between different features, and the usefulness of such systems was limited due to the difficulty in representing high-level concepts using low-level feature. This paper presents a novel neural network which is based on the nonlinear kernel Least Mean Square (KLMS). The proposed approach allows the users to select an initial query image and incrementally search target images via relevance feedback. If users arenʹt satisfied with the retrieved results, relevance feedback method is used to enhance the performance of the proposed system by updating a boundary for separating relevant images from irrelevant ones. These two similarity measures are normalized and combined together to form the overall similarity measure. Experimental results using a COREL database demonstrate the effectiveness of the proposed method