A Bayesian Framework for Recognizing Textured Objects in a Content- Based Image Retrieval System

We present an image retrieval system for finding textured objects by text query. The algorithm considers each image as a set of independent segments. It learns the relationship of segments features with text from a training set of images where a set of segments is manually labeled. The algorithm is capable of generating text labels from a segment and finding images that are relevant to a query consisting of text labels. We present the first experimental results.