SGER: Developing a Non-parametric Digital Image Search Engine

点击翻译按钮获取中文摘要

*** This award, in the Small Grants for Exploratory Research mode, will investigate ways to group images to facilitate the development of effective indexing methods for image search. Most current approaches to this problem focus on texture and color characteristics in the images. This effort, building on prior funded research in this laboratory on nonparametric shape modeling, seeks to determine whether structural encoding can be used more effectively to give the search engine more information about the image, and thereby facilitate the search. Generally, contemporary approaches have avoided relying on such structural information as too difficult to obtain. The planned approach will combine geometrical and topological considerations such as symmetry, parallelness, orthogonality, and connectivity to gainmore of an understanding of the image content, leading to the production of digital geometric primitives to aid in image search. These geometric/topologic primitives should complement the texture and color information currently used for this purpose.***