Performance Evaluation of Parallel Image Compression Using DCT and Huffman with OpenMP and Intel TBB
Annotatsiya
In this study, image compression was performed using DCT and Huffman coding processes, which are widely used for image compression. Since this process is image-based, it runs very slowly on devices without a graphics processing unit (GPU). The goal is to organize parallel calculations on the central processing unit (CPU). The study initially studies the theoretical foundations of the compression algorithm and the principles of operation of DCT (Discrete Cosine Transform) and Huffman coding on block-based images. The study also analyzes approaches to parallelization in image compression processes, studies the OpenMP (Open Multi-Processing) and Intel TBB (Threading Building Blocks) architectures designed for parallel processing on processors, their load balancing, synchronization and scheduling mechanisms, and implements parallel image compression processes using these tools. As an experiment, a small <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$256 \times 256$</tex> image is used in comparison with a large <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4 0 9 6 x} \boldsymbol{4 0 9 6}$</tex> image. Based on the research results, it was found that the OpenMP parallel computing model is more effective in compressing small-sized images <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(256 \times 256,512 \times 512)$</tex> due to its simplicity, while the Intel TBB-based model provides higher speed for the remaining largesized images. At the end of the study, using the evaluation indicators (including speed, efficiency, and compression quality) of these two parallel computing models for image compression, it was found that parallel programming is an effective approach to optimize image compression processes using parallel computing models such as OpenMP and Intel TBB on non-GPU devices.