# Software optimization of Discrete Cosine (or Sine) Transform for Versatile Video Coding _by Mehat Baptiste (IETR-VAADER) - 2019.08.20_ ###### tags: `VAADER` `Seminar` ![](https://i.imgur.com/6KG2KDC.png) ## Abstract As display technology evolves, video quality improves and video files get bulky. Achieving real-time processing on such big files requires high computational power. Video compression is then used in order to reduce the amount of data to be processed without significant information loss. The current video compression standard, HEVC (High Efficiency Video Coding), partitions individual video pictures into pixel blocks whose size can vary into a restricted range. These blocks are then processed to achieve the compression. The future compression standard, VVC (for Versatile Video Coding), introduces a wider range of sizes for these blocks. This implies that algorithms processing these have to be re-adapted and optimized for the new block sizes. One of these algorithms performs Discrete Cosine (or Sine) Transforms (DCT or DST) on blocks to reveal spatial redundancies. This part of the program mainly consists of matrix multiplication and could be optimized to reduce its time cost. This talk will present how this optimization could be done using Intel's vectorial extensions such as SSE2 and AVX2. ## [Slides](https://mycore.core-cloud.net/index.php/s/0qHxkHrw3UrFmzA)