Analysis of parallel multicore performance on sobel edge detector

  • Authors:
  • Noor Elaiza Abdul Khalid;Siti Arpah Ahmad;Noorhayati Mohamed Noor;Ahmad Firdaus Ahmad Fadzil;Mohd Nasir Taib

  • Affiliations:
  • Faculty of Computer Science and Mathematics, Faculty of Electrical and Electronic Engineering, University Technology MARA, Shah Alam, Selangor, Malaysia;Faculty of Computer Science and Mathematics, Faculty of Electrical and Electronic Engineering, University Technology MARA, Shah Alam, Selangor, Malaysia;Faculty of Computer Science and Mathematics, Faculty of Electrical and Electronic Engineering, University Technology MARA, Shah Alam, Selangor, Malaysia;Faculty of Computer Science and Mathematics, Faculty of Electrical and Electronic Engineering, University Technology MARA, Shah Alam, Selangor, Malaysia;Faculty of Computer Science and Mathematics, Faculty of Electrical and Electronic Engineering, University Technology MARA, Shah Alam, Selangor, Malaysia

  • Venue:
  • Proceedings of the 15th WSEAS international conference on Computers
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper presents the parallel multicore Sobel edge algorithm which parallelizes the traditional sequential Sobel edge detection algorithm on a parallel multicore platform. The current advancement of multicore architecture can be utilized by the parallel programming paradigm when focuses on the thread operations. The CPUs/cores provide more processing resource to be used but often not fully utilized to its utmost potential. In this paper, the performance of multicore architectures, combined with the parallel communication software named Message Passing Interface (MPI), on the application of Sobel Edge detector algorithm is implemented on various thread processing is performed and analyzed. The test is being done on ten different images with each image tested in the varying size of 1K×1K, 2K×2K, and 3K×3K pixels. The significant performance increment is discovered due to the fact that the CPU is being fully utilized. This proves that the current hardware is far more underutilized than one would expect.