Multi-Threaded Computation of the Sobel Image Gradient on Intel Multi-Core Processors Using OpenMP Library

Performance of applications executed on multi-core processors is not boosted by just dividing the work among a team of threads and assigning them blindly to the CPU cores. Factors such as data access patterns in memory, the way of allocating the threads to the physical cores, and how the data are partitioned among the threads significantly affect performance. In this paper, we target the acceleration of the Sobel image gradient computing which is important in segmenting images for further processing in computer vision and image analysis applications. We present a multi-threaded algorithm using the standard OpenMP threading library to parallelize the computations using two Intel multi-core processors. The effects of the parallelization factors on the performance of the proposed algorithm are evaluated using different image resolutions to draw accurate conclusions. Our results showed a maximum attained speedup closer to the number of physical cores in the CPU, which is the maximum theoretical value.