CBIR using content frequency and colour features

Abstract

Due to the diversity of the image content, we propose in this study a new technique for content-based image retrieval (CBIR) to characterise the image. In this scenario, all images are characterised by their frequency content and their colour information. Indeed, using the high resolution spectral analysis methods, especially the 2D estimation of signal parameters via rotational invariance techniques (ESPRIT) we extract from the image its content frequency and with statistical moment its colour information in order to construct a new vector descriptor. The experimental results applied to the coil-100 database show the robustness of our approach, the precision average can be reaches 90.57%.