In this thesis, we investigate the effectiveness of semantic fisheye views for opportunistic search and sensemaking tasks over large collections of annotated images. Semantic fisheye views are interactive visualization techniques that reveal semantic detail and context related to a focus. These techniques reduce visual complexity by showing the most important information in detail, and de-emphasizing or filtering less important information. There are three significant contributions of this thesis. First, we have developed a framework for the design of semantic fisheye view techniques to support highly interactive tasks, such as opportunistic search and sensemaking. Second, we have developed a prototype for exploring a large collection of annotated images based on this framework. The workspace integrates WordNet to support semantic queries and navigation over the unstructured image collection. Finally, we present the results of an experimental evaluation comparing the effectiveness of similarity and semantic-guided search for a complex search task. The results of this experiment strongly suggest that semantic-guided search is more effective for opportunistic search and sensemaking tasks.