A Distributed Algorithm for Multispectral Image Segmentation

Conference Paper · October 2007with17 Reads
DOI: 10.1109/SYNASC.2007.33 · Source: IEEE Xplore
Conference: Symbolic and Numeric Algorithms for Scientific Computing, 2007. SYNASC. International Symposium on

    Abstract

    This paper presents a distributed algorithm for multi-spectral image segmentation. Regions of the image are processed separately and then the results are combined. For this, the algorithm employs two types of clustering algorithms, each specialized in its task and steered toward obtaining a final meaningful segmentation. The workers use an iterative clustering algorithm which is an extension of fuzzy c-means that uses spatial information and a spectral compatibility heuristic. An agglomerative clustering algorithm is used by the master to combine the partial results. The algorithm is used for the segmentation of satellite images coming from the MODIS sensor aboard the Terra and Aqua satellites. The quality of the segmentation results and the speedups obtained by using the grid are discussed.