Application of Fuzzy Edge Detection for Fast Object-based Image Retrieval

This paper describes our use of fuzzy logic techniques for fast object-based retrieval of images. To start with, the user specifies a region containing the object of interest. The image processing methods that are employed work with gray level versions of the region image and of the images in the database. In the retrieval process, we use edge based representations. Fuzzy edge detection coupled with edge thinning forms the primary process of creating edge-maps. Images containing the object are retrieved by similarity matching using a multi-resolution feature list consisting of edge crossings with grid lines. For speed, edge detected multi-resolution representations are computed in a preprocessing operation for all images in the database. Experiments on a database of images obtained using a simple camera yielded good retrieval performance