A comparative study of feature vectors derived from wavelets applied to high resolution satellite images retrieval

Nowadays, content-based image-retrieval techniques constitute powerful tools for archiving and mining of large remote sensing image databases. High spatial resolution images are complex and differ widely in their content, even in the same category. All images are more or less textured and structured. If the image to recognize is somewhat or very structured, a shape feature will be somewhat or very effective. As if the image is composed of a single texture, a parameter reflecting the texture of the image will reveal more efficient. It this paper, we propose to compare different feature vectors extracted from wavelet decomposed remote sensed images as wavelet is ideally suited for highlighting local feature points in the decomposed subimages. Using approximation and details images, we computed several features namely statistical moments, Zernike moments, Histograms of Gradients (HOG), color histograms, Local Binary Pattern (LBP) histograms and compared their performances in term of precision and recall. The preliminary results show that statistical moments outperform the other features. They permit to reach more accuracy and better performances regarding to retrieval results and time computation.