Browse Theses

Recent technological advances have made it possible to process and store large amounts of image/video data. Perhaps the most impressive example is the fast accumulation of image data in scientific applications such as medical and satellite imagery. The internet is another excellent example of a distributed database containing several millions of images. However, in order to realize their full potential, tools for automated analysis and extraction of information, and for intelligent searches in image databases need to be developed.

We have investigated various techniques which facilitate content-based image search and retrieval. A prototype system, called NETRA, which enables the search of aerial photographs and natural color images has been implemented on the web using the platform independent Java language. A distinguishing aspect of this system is its incorporation of a robust automated image segmentation algorithm that allows object or region based search. Image segmentation significantly improves the quality of image retrieval when images contain multiple complex objects. Images are segmented into homogeneous regions at the time of ingest into the database, and image attributes that represent each of these regions are computed. This is the first time that image segmentation and region based search have been combined in a robust way and retrieval performance demonstrated on a large image database.

In addition to image segmentation, other important components of the system include feature representations for characterizing the color, texture, and shape information, an approach to enhancing the retrieval performance by learning the appropriate similarity measures in the image feature space, and an image thesaurus model for image annotation and indexing. NETRA allows users to search by image example. For instance, the user can retrieve all images containing "blue sky" by specifying the color (blue) and location (upper one-third) information. Images containing snow covered peaks can be specified by selecting an example from the database and choosing color and texture attributes for search. NETRA can be accessed on the web at "http://vivaldi.ece.ucsb.edu/Netra."