Xiumei Li (M.Sc.)

My research focuses on learning-based scalable point cloud compression. Compared with 2D images, point clouds provide richer spatial information and a more immersive experience, making them widely used in virtual/augmented reality, industrial robotics, and autonomous driving. Due to their high data volume, efficient compression is essential for storage and transmission. While traditional hybrid point cloud compression methods rely on hand-crafted priors that limit compression performance, recent deep learning-based approaches have achieved significant improvements. However, most existing learned PCC methods are designed for a single quality level, making them ill-suited for practical scenarios that require flexible bitrate adaptation and multi-quality reconstruction. My work addresses this by developing scalable learned point cloud compression methods that support efficient decoding across multiple quality levels from a single bitstream.