A Picture Retrieval System Based on Contents

点击翻译按钮获取中文摘要

This is an inter-institutional collaborative research project involving University of Illinois at Chicago (UIC) and Wright State University (WSU). Pictorial systems which retrieve pictures based on contents have wide applications. For example, learning can be facilitated by retrieving pictures from electronic encyclopedia; radiologists can be assi sted in diagnosis; satellite images of desired contents can be retrieved by earth scientists to perform research; appropriate garments can be viewed by customers before deciding where they go shopping, etc. The purpose of the project is to construct a generic picture retrieval and analysis system which can be easily adapted to various applications , including the construction of digital libraries and their use in retrieval of information. Similarity based retrieval involving spatial relationships, and fuzzy and precise matching is used one of the approaches investigated. Other techniques that play a significant role are: multi-dimensional searching; parametrized similarity functions; visual query tools; image analysis techniques such as segmentation; object recognition; and color and shape analysis. The retrieval subsystem is developed at UIC, while the image analysis subsystem is developed at WSU. The two subsystems will be combined to form an integrated system. The usefulness of the system will be demonstrated for the following ap plications: retrieval of photographs for the photo-shop at UIC, and retrieval of images for diagnosis of skin cancer at Rush-Presbyterian Medical Center in Chicago.