Engineering an image‐browsing environment: re‐purposing existing denotative descriptors

The paper provides a theory base for deriving connotative descriptors for photographs from existing denotative descriptors, and then demonstrates a model for enhancing browsing within image collections by providing a tool for carving up the searching space.

The paper conceptually explores the nature of iconic messages contained in an image by adopting semiotics as a theoretical tool. A problem of image retrieval is identified as loss of connotative messages during the image representation process. The paper proposes an image‐retrieval model utilizing an association thesaurus that facilitates the assignment of connotative index terms by making use of denotative index terms of an image. A series of experiments are performed for evaluating the effectiveness of the proposed model.

Experimental results demonstrate that the association thesaurus improves image‐retrieval effectiveness by increasing the recall of connotatively related image documents as well as the recall of browsing sets.

Applying connotative index terms to an image would be time consuming. Deriving connotative terms from denotative terms and then using them to enrich the browsing environment suggest a method of increasing retrieval effectiveness while reducing the resources required for representation.

Since images are often used to illustrate concepts that are not immediately evident from just the objects in front of the lens, connotative descriptions are particularly valuable. Since human perception of images is, in a sense, hard wired into our brains, browsing is a frequent and reasonable search method in image collections. Using connotative descriptors to point the way to clusters of images with a higher probability of relevance changes the locus of control over representation establishes an environment for dynamic representation, and gives credibility to browsing as a significant search method.