一种面向多核处理器的高效并行PCA-SIFT算法

摘要

An efficient parallel PCA-SIFT algorithm for multi-core processor is proposed, which adopts data-level parallel method to implement parallel key points extraction and matching. The computing tasks were assigned to four DSP cores for parallel processing, which fully develops multi-level parallelism of multi-core processor. The experimental results show that the parallel PCA-StFT algorithm has good adaptability for various distortion images and the image matching ability is close to the sequential algorithm with the average speedup 3.12.%提出一种面向多核处理器的并行PCA-SIFT算法,采用数据级并行方法实现并行的特征提取和特征点匹配,将计算任务分配到各个DSP核并行处理,充分开发多核处理器的多级并行性.实验结果表明,并行PCA-SIFT算法对各种不同图像形变的图像具有良好的适应性,具有接近串行PCA-SIFT算法的图像匹配能力,平均加速比达3.12.