An Efficient CBIR System for High Resolution Remote Sensing Images
High resolution satellite images (HRSI), contain a great range of objects and spatial patterns appearing with a large variation of scale, rotation and illumination. In this paper, we propose a Content Based Image Retrieval (CBIR) system for HRSI by employing SURF features and boost them through the addition of color, texture and structural information around key points; and by learning a category-specific dictionary for each image class. An extensive experimental evaluation on the well- known UC Merced dataset has been performed and compared with other feature extraction methods including Convolutional Neural Networks. It is demonstrated that our method is quite competitive in terms of performance.
KeywordsCBIR Retrieval SURF BOVW High resolution satellite images
- 4.Yang, X., Wen, G.: Road extraction from high-resolution remote sensing images using wavelet transform and hough transform. In: 5th International Congress on Image and Signal Processing, China (2012)Google Scholar
- 5.Chen, C., et al.: Gabor-filtering-based completed local binary patterns for land-use scene classification. In: IEEE International Conference on Multimedia Big Data (BigMM), China (2015)Google Scholar
- 6.Bouteldja, S., Kourgli, A.: Multiscale texture features for the retrieval of high resolution satellite images. In: International Conference on Systems, Signal and Image Processing (IWSSIP), London (2015)Google Scholar
- 7.Pham, M.T., et al.: Texture retrieval from VHR optical remote sensed images using the local extrema descriptor with application to vineyard parcel detection (2016). https://hal.archives-ouvertes.fr/hal-01311993
- 13.Erchan, A.: Bag of morphological words for content-based geographical retrieval. In: Proceedings of the International Workshop on Content-Based Multimedia Indexing, Austria, pp. 1–5 (2014). https://doi.org/10.1109/cbmi.2014.6849837