Scholarlink

This study explores the parallelization of the Local Triangular Coding Pattern (LTCP) descriptor using the Message Passing Interface (MPI) to enhance computational efficiency in image processing. By distributing workload across multiple processes, the MPI-based implementation achieves substantial reductions in execution time, with speedup varying across different image sizes. Experimental results show a peak speedup of 4.8899 under optimized compiler settings, demonstrating the effectiveness of MPI in accelerating LTCP computation. Compared to alternative parallelization techniques, MPI excels in distributed environments but faces challenges such as inter-process communication overhead and diminishing returns beyond a certain number of processes. Scalability and efficiency analyses further highlight its strengths and limitations, providing insights into optimal resource allocation. The proposed approach is particularly beneficial for real-time image analysis applications, including medical imaging, remote sensing, and object recognition. Future work will explore hybrid models combining MPI with shared-memory paradigms to further optimize performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Implementation of Local Triangular Coded Pattern Using MPI Parallel Framework

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shrutha V. Bhat,
  • Rakshith H. Kalmadi,
  • B. Ashwath Rao,
  • N. Gopalakrishna Kini

摘要

This study explores the parallelization of the Local Triangular Coding Pattern (LTCP) descriptor using the Message Passing Interface (MPI) to enhance computational efficiency in image processing. By distributing workload across multiple processes, the MPI-based implementation achieves substantial reductions in execution time, with speedup varying across different image sizes. Experimental results show a peak speedup of 4.8899 under optimized compiler settings, demonstrating the effectiveness of MPI in accelerating LTCP computation. Compared to alternative parallelization techniques, MPI excels in distributed environments but faces challenges such as inter-process communication overhead and diminishing returns beyond a certain number of processes. Scalability and efficiency analyses further highlight its strengths and limitations, providing insights into optimal resource allocation. The proposed approach is particularly beneficial for real-time image analysis applications, including medical imaging, remote sensing, and object recognition. Future work will explore hybrid models combining MPI with shared-memory paradigms to further optimize performance.