An open source agenda for research linking text and image content features

AbstractThe use of primitive content features of images for classification and retrieval has matured over the past decade. However, human beings often prefer to locate images using words. This article proposes a number of methods to utilize image primitives to support term assignment for image classification. Further, the authors propose to release code for image analysis in a common tool set for other researchers to use. Of particular interest to the authors is the expansion of work by researchers in image indexing to include image content based feature extraction capabilities in their work.