Image Mining: Improving Image Retrieval Technique Using Variational Models and Relevance Feedback

Recently, multimedia data are growing rapidly in our life. Images have been playing a vital role in many fields and carry valuable information. This information needs to be processed. In addition, image mining field has seemed to present a valuable help to systems and persons who need to extract knowledge from images dataset. Content-Based Image Retrieval (CBIR), a type of image mining, aims to retrieve similar images from large image databases according to users' interests. However, the big problem with this technique is how to recover pertinent images. In this work, we focus on improving the performance of Content-Based Image Retrieval systems by reducing the semantic gap between query images and the similar retrieved images. The proposed approach gives an effective solution based on TV-L1 model and relevance feedback in order to ensure a relevant searching. Finally, the evaluation is made using the precision and recall metrics to demonstrate the results accuracy.