Abstract: In this paper, we present an efficient progressive out-of-core compression algorithm, called POoCC, for the gigantic models that already exceed the capacity of physical main memory. To our knowledge, the scheme is the first solution of progressive out-of-core encoding. The primary obstacle of extending current in-core progressive encoders to their out-of-core versions is that the necessary global data structure cannot be built and kept in-core for gigantic meshes. We partition the model space into small blocks and use a multi-level adaptive tree to organize the whole input gigantic mesh. Using the multi-level adaptive tree, most of the progressive encoders based on tree structure can be easily adapted to deal with gigantic meshes. Because of its high performance, we choose to extend the in-core encoder for complex isosurfaces [Lee et al. 2003] to handle gigantic meshes. Our POoCC algorithm achieves the similar rate distortion performance as the state-of-the-art in-core encoders and performs better than the existing single rate out-of-core compression algorithms, especially for those watertight gigantic meshes. The gigantic meshes are reconstructed using the Extended Marching Cubes (EMC) algorithm[Kobbelt et al. 2001]. Thus, they can be progressively compressed in one piece and no artificial discontinuity is introduced even the partitioning procedure is necessary, which makes our POoCC encoder outperform other model splitting based out-of-core processing algorithms. POoCC is carefully designed to be cache-friendly and can handle meshes with arbitrary size on low-cost off-the-shelf PCs.