The pattern problem is important in image retrieval because its solution may allow users to query a database for images containing a specific item (e.g. object, product logo, machine part, etc.). This is a very common form of query in a text-based system, where, for example, a user may be looking for all articles that mention a specific term.
The main difficultly in efficiently solving a single pattern problem is the combination of partial matching and scaling. Allowing partial matching means that we do not know where in the image to look, and allowing scaling means that we do not know how much of the image to look at. Furthermore, there are many independent (i.e. non-overlapping) places in the image to look for very small scale occurrences of a pattern. In addition to these efficiency difficulties, there is the fundamental issue of how to combine color (shape) information with position information in judging the similarity of two color (shape) patterns.
In the context of database
query, the fact that partial match distance measures do
|top||Title, Table of Contents, Introduction|
|next||The SEDL Image Retrieval System|
The ideas and results contained in this document are part of my thesis, which will be published as a Stanford computer science technical report in June 1999.
S. Cohen. Finding Color and Shape Patterns in Images. Thesis Technical Report STAN-CS-TR-99-?. To be published June 1999.
Email comments to firstname.lastname@example.org.