A programming model for distributed content-based image retrieval

Abstract

With the growth of digital contents, content-based image retrieval becomes important. The huge amount of computation can be solved by distributed computing. However, to design a distributed algorithm and program it with cross-platform capability is difficult. We present a programming model which is based on list data processing, and it can be transparently executed in a distributed environment with some auxiliary tools. This model allows us to explicitly specify the parallelism without special language support, such as new reserved keywords or new directives. Although this approach may introduce non-trivial overhead in some aspects, it can be used to write a distributed program using a conventional (and often sequential) language.