A Study of Connected Component Labeling Algorithms on the MPP
نویسندگان
چکیده
In this paper we consider the problem of labeling the connected components of a binary image. We descnbe three different algorithms and variations thereof which we implemented on NASA's MPP. These algoritluns include a simple label propagation algorithm. a version of Levialdi's algorithm, and an asymptotically optimal divide-and-eonquer algorithm. We discuss the perfonnance of these algorithms on the MPP and provide insight into how special hardware features of the lvIPP influence the design of parallel algorithms. We also address the issue of how to handle images that are larger than the processor array of the MPP (Le.; images larger !ban 128xl28). This work was supponed by the Office of Naval Research under Contracts NOOOl4-84-K-OS02 and NOOO14-86.K-0689, and by the National Science Foundation lDldcr Gn!nt DMC.84-13496.
منابع مشابه
In-database connected component analysis
We describe a Big Data-practical, SQL-implementable algorithm for efficiently determining connected components for graph data stored in a Massively Parallel Processing (MPP) relational database. The algorithm described is a linear-space, randomised algorithm, always terminating with the correct answer but subject to a stochastic running time, such that for any ǫ > 0 and any input graph G = 〈V,E...
متن کاملParallel Execution of a Connected Component Labeling Operation on a Linear Array Architecture
This work presents a novel parallel algorithm and architecture for finding connected components in an image. Simulation results indicate that the proposed algorithm has an execution time of N+6N-4 cycles for an N×N image using an architecture containing 4 parallel processors. The proposed hardware can process a 128 × 128 image in 0.8574 ms and uses only 4 processors, compared to 0.85 ms and 128...
متن کاملConnected Component Labeling Algorithms for Gray-Scale Images and Evaluation of Performance using Digital Mammograms
The main goal of this paper is to compare performance of connected component labeling algorithms on grayscale digital mammograms. This study was carried out as a part of a research for improving efficiency and accuracy of diagnosing breast cancer using digital mammograms. Three connected component labeling algorithms developed by Jung-Me Park [8], Kenji Suzuki [16] and Kesheng Wu [9], were used...
متن کاملTwo Strategies to Speed up Connected Component Labeling Algorithms
This paper presents two new strategies to speed up connected component labeling algorithms. The first strategy employs a decision tree to minimize the work performed in the scanning phase of connected component labeling algorithms. The second strategy uses a simplified union-find data structure to represent the equivalence information among the labels. For 8-connected components in a two-dimens...
متن کاملAn Improved Transformerless Grid-Connected PV System With Related Control Strategy to Reduce Leakage Current, Extract Maximum Power Point and Control Power
The main purposes of a transformerless grid-connected photovoltaic (PV) system consist of the reduction of leakage current, extraction of maximum power point (MPP), tracking of MPP (MPPT), controlling the active and reactive powers, and having the unity power factor. To achieve the above-mentioned aims, the following actions have been performed in this paper. First of all, a brief analysis of t...
متن کامل