Chapter 4: Clustering and Classification

Title: Multispectral image indexing using a joint approach in the spatial and the compressed domains

Author(s): Rafik Bensalma, Mohamed-Chaker Larabi, Christine Fernandez-Maloigne

Address: SIC Laboratory EA 4103, University of Poitiers, France | SIC Laboratory EA 4103, University of Poitiers, France | SIC Laboratory EA 4103, University of Poitiers, France

Reference: Atlantic Europe Conference on Remote Imaging and Spectroscopy pp. 102 - 109

Abstract/Summary: This work lies within the scope of content-based image retrieval and browsing. Our interest is related on the one hand, to the extraction of features in the compressed domain with the JPEG2000 standard and on the other hand, to the multispectral images acquired within the framework of project PIMHAI (European project INTERREG III B ATLANTIC AREA) dedicated to the monitoring of the environment of the Atlantic littoral. Let us recall that the multispectral images are characterized by a high number of bands and an important resolution. In order to extract the relevant content from the images and then to build the signatures which will represent them, we jointly used features calculated in the spatial domain and others in the compressed domain. These latter features are applied to all the subbands of the wavelet transform (DWT) except for low frequency subband (named LL). This one is represented using features from the spatial domain. The similarity measurement between images is calculated using the fusion of the features carried out by the means of weighting factors resulting from a preliminary training. The results are posted on a browsing interface, which is installed in order to simplify the navigation and retrieval.

Order a copy of this article