Host: The Institute of Image Electronics Engineers of Japan
Co-host: The Institute of Image Information and Television Engineers
This work generalizes the concept of blobs defined in the scale-space theory to contain color information. We also propose an image categorization algorithm based on those color blobs. Our approach relies on color quantization using basic color categories and the successive classification of the resulting regions using neural networks. We show that the preattentive information can be used to categorize even low resolution pictures taken with mobile devices. Thus, we propose a framework in a mobile environment where the classification and additional contextual information, such as GPS location, are used to assist user in such tasks as navigation or self-configuration of the device.