Mobile Image Retrieval using Morphological Color Segmentation

An image retrieval method targeted to mobile devices is proposed in this paper. The method is based on an efficient segmentation method that obtains the most significant regions in the picture. These regions are found using a neural networkbased color quantization over the morphological scale space of the images. The features of each segment are then used to index images in a database. The images also contain physical location context, by means of GPS, and the retrieval system is used as a part of an intelligent mobile interface. The retrieval method is compared with methods based on color histograms, showing the increase in speed, recall and precision, while maintaining interactive speeds in mobile devices.