Furthermore, the syntax generality does not require major program rewritings to be able to describe the parallelism. on Image Processing, 1996, 299-302 [10] Shirazi, B., Wang, M.: Analysis and Evaluation of Heuristic Methods for Static Task Scheduling. Many fast block-matching algorithms for motion estimation have been proposed because of their lower computation overhead than that of full- search block-matching algorithm, such as the existing three-step search (TSS) algorithm [1] and the recently proposed new three-step search (NTSS) algorithm [2].

However, when the application presents more complex parallelism patterns, per- formances can be increased up to 12% for the LU SPLASH benchmark compared with classical parallelism management. 10, 1990, 222-232 [11] Park, G.-L., Shirazi, B., Marquis J., Choo, H.: Decisive Path Scheduling: A New List Scheduling Method. 1 Introduction It is well known that motion estimation algorithms play an important role in video sequence compression.

This conference is a merge of two established international conference/workshop series devoted to parallel processing: the ACPC conferences which were held previously in Salzburg, Gmunden, and Klagenfurt (all Austria) and the Parallel Numerics (Par Num) workshops which were organized in Smolenice (Slovakia), Sorrento (Italy), Gozd Martuljek (Slovenia), and Zakopane (Poland). The mean absolute error (MAE) is used as an appropriate estimate of the block distortion measure (BDM).

The partners of ACPC run research projects in these fields, use a common pool of hardware equipment, and offer a joint curriculum in Parallel Computation for graduate and postgraduate students. Additionally, the ITSS employs a smaller number of search points than TSS to speed up block matching.

Moreover, sev- eral national and international workshops and conferences have been organized within the framework of the ACPC. The details of the algorithm are given in the example described below.

Indeed, the parallelism is no more spread throughout the program code and is decoded in parallel to task execu- tion.

We thank all reviewers listed below for their valuable work. The separate parallelism description allows to implement a high pattern complexity with a limited dynamic overhead.