Although many image processing applications are ideally suited for parallel implementation, most researchers in imaging do not benefit from high performance computing on a daily basis. Essentially, this is due to the fact that no parallelization tools exist that truly match the image processing researcher's frame of reference. As it is unrealistic to expect imaging researchers to become experts in parallel computing, tools must be provided to allow them to develop high performance applications in a highly familiar manner. In an attempt to provide such a tool, we have designed a software architecture that allows transparent (i.e., sequential) implementation of data parallel imaging applications for execution on homogeneous distributed memory MIMD-style multicomputers. This paper presents an extensive overview of the design rationale behind the software architecture, and gives an assessment of the architecture's e#ectiveness in providing significant performance gains. In particular, we describe the implementation and automatic parallelization of three well-known example applications that contain many fundamental imaging operations: (1) template matching, (2) multi-baseline stereo vision, and (3) line detection. Based on experimental results we conclude that our software architecture constitutes a powerful and user-friendly tool for obtaining high performance in many important image processing research areas.

A popular approach to providing non-experts in parallel computing with an easy-to-use programming model, is to design a software library consisting of a set of pre-parallelized routines, and hide the intricacies of parallelization behind the library's API. However, for regular domain problems (such as simple matrix manipulations or low level image processing applications | in which all elements in a regular subset of a dense data eld are accessed in turn) speedup obtained with many such library-based parallelization tools is often sub-optimal. This is because inter-operation optimization (or: time-optimization of communication steps across library calls) is generally not incorporated in the library implementations.

The Parallel-Horus framework, developed at the University of Amsterdam, is a unique software architecture that allows non-expert parallel programmers to develop fully sequential multimedia applications for efficient execution on homogeneous Beowulf-type commodity clusters. Previously obtained results for realistic, but relatively small-sized applications have shown the feasibility of the Parallel-Horus approach, with parallel performance consistently being found to be optimal with respect to the abstraction level of message passing programs. In this paper we discuss the most serious challenge Parallel-Horus has had to deal with so far: the processing of over 184 hours of video included in the 2004 NIST TRECVID evaluation, i.e. the de facto international standard benchmark for content-based video retrieval. Our results and experiences confirm that Parallel-Horus is a very powerful support-tool for state-of-the-art research and applications in multimedia processing.

Performance obtained with existing library-based parallelization tools for implementing high performance image processing applications is often sub-optimal. This is because inter-operation optimization (or: optimization across library calls) is often not incorporated in the library implementations. This paper presents a simple, efficient, finite state machine-based method for global performance optimization, called 'lazy parallelization'. Experimental results based on this approach show significant performance improvements over non-optimized parallel implementations.

In the research area of image and video processing, the Horus library [5] has become a valuable system for software driven rapid prototyping, and an essential vehicle for knowledge transfer at the level of universities and expertise centers. Due to its strictly sequential implementation, however, Horus can not always satisfy all demands for high performance. As computational...

One of the most fundamental tasks any automatic parallelization and optimization tool is confronted with is to find an optimal domain decomposition for an application at hand. ForregBMB domain problems (such as simple matrix manipulations) this task may seem trivial. However, communication costs inmessag passing progng oftensignfiR cantly depend on the capabilities and particular behavior of the applied communication primitives. As a consequence, straiguence,fiR domain decompositions may deliver non-optimal performance.

Multimedia data is rapidly gaining importance along with recent developments such as the increasing deployment of surveillance cameras in public locations. In a few years time, analyzing the content of multimedia data will be a problem of phenomenal proportions, as digital video may produce data at rates beyond 100 Mb/s, and multimedia archives steadily run into Petabytes of storage space. Consequently, for urgent problems in multimedia content analysis, Grid computing is rapidly becoming indispensable. This demonstration shows the viability of wide-area Grid systems in adhering to the heavy demands of a real-time task in multimedia content analysis. Specifically, we show the application of a Sony Aibo robot dog, capable of recognizing objects from a set of learned objects, while connected to a large-scale Grid system comprising of cluster computers located in Europe, the United States, and Australia. As such, we demonstrate the effective integration of state-ofthe-art results from two largely distinct research fields: multimedia content analysis and Grid computing. 1.