Parallel Algorithm for Concurrent Computation of Connected Component Tree P. Matas 1,2 , E. Dokladalova 1 , M. Akil 1 , T. Grandpierre 1 , L. Najman 1 , M. Poupa 2 , and V. Georgiev 2 1 IGM, Unité Mixte CNRS-UMLV-ESIEE UMR8049, Université Paris-Est, Cité Descartes, BP99, 93162 Noisy le Grand, France {matasp, e.dokladalova, akilm, grandpit, l.najman}@esiee.fr 2 Department of Applied Electronics and Telecommunications, University of West Bohemia, Univerzitní 26, 306 14 Plzeň, Czech Republic {pmatas, poupa, georg}@kae.zcu.cz Abstract. The paper proposes a new parallel connected-component-tree construction algorithm based on line independent building and progressive merging of partial 1-D trees. Two parallelization strategies were developed: the parallelism maximization strategy, which balances the workload of the processes, and the communication minimization strategy, which minimizes communication among the processes. The new algorithm is able to process any pixel data type, thanks to not using a hierarchical queue. The algorithm needs only the input and output buffers and a small stack. A speedup of 3.57 compared to the sequential algorithm was obtained on Opteron 4-core shared memory ccNUMA architecture. Performance comparison with existing state of the art is also discussed. 1 Introduction Computer vision systems are asked to furnish high performance and be flexible for a large variety of existing or possible applications. One of the global problems of the vision system design is how to achieve these two characteristics simultaneously. If the high performance is achieved by an optimization effort which means a kind of system specialization, it will (by definition) limit its flexibility. The connected component tree (CCT) based image processing algorithms seem to be very promising from this point of view. They allow bridging the gap between low- and high-level processing implementations. They have been used for filtering [1, 5] as well as the image analysis: motion extraction [1], watershed segmentation [6, 2, 14], segmentation of astronomical images [3] or data visualization [13]. Fig. 1 shows typical stages of an application based on CCT. We can see the advantage of these methods: once the CCT is constructed, the processing is performed on the tree by graph transformation(s), and only one data structure is used from low- level to high-level processing. In addition, the graph transformations are applicable to any dimension (1D, 2D, 3D …).