eb content-based image retrieval is being implemented in many databases, especially in digital libraries. In many specialized fields, professionals often wish the content-based image retrieval system could provide more sensitive shape and color matching to exploit or classify some kinds of species. In our experience, ecosystem synopsis is the obvious illustration. Hence, we offer a series of methods to solve these problems: the defined signature of the object, with object detection/separation and normalization; the shape representations and similarity measurement with affine invariants; the color feature and similarity measurement; and the content-based image retrieval system requirements and design issues.
First international workshop on software engineering and multimedia applications(SEMA'99), Baden-Baden, Germany