Skip to Main Content
Relevance feedback has drawn intense interest from many researchers in the field of content-based image retrieval (CBIR). In recent years, kernel-based approach has been a popular choice for the implementation of the relevance feedback based CBIR system. This is largely due to its ability to classify patterns with limited sample data. Since most of the kernel approaches reported have been treating the input as a long flat vector, such arrangement may increase the chances of “polluting” the feature element that uniquely identifies the selected image group. This paper proposes a two layer kernel configuration with an objective to improve the retrieval accuracy. While the performance of the two configurations is similar in certain conditions, the proposed configuration has shown to superior when dominant feature element exists that is capable to uniquely identify the selected image group.