Existing content-based image retrieval paradigms almost never address the problem of starting the search, when the user has no starting example image but rather a mental image.

Many current content-based image retrieval systems suffer from poor result presentation. A sophisticated visualization can be used to identify differences between human perception of similarity and system-determined similarity. Analyzing such discrepancies is a prerequisite for a system trimmed towards the user’s comprehension of the underlying retrieval process. The aim of the visualization techniques presented in this paper is to provide a tool to analyze a mismatch between the user’s perception and the system’s calculation of similarity. We combine techniques of visual image retrieval and information visualization to acquire insight into the extracted feature data. In our project we implemented visualization techniques to present feature data on three different levels of abstraction. We discuss our experiences when working with a Data Table, a Parallel Coordinate Plot, and a Color Space Plot. 1.

the relevance feedback mechanism

Abstract. In this paper an application of the TS-SOM variant of the self-organising map algorithm on the problem of copyright theft detection for bitmap images is shown. The algorithm facilitates the location of originals of copied, damaged or modified images within a database of hundreds of thousands of stock images. The method is shown to outperform binary decision tree indexing with invariant frame detection.