Abstract: In order to solve computer storage and transmission problem due to massive 3D point cloud,
a fast and lossy compression algorithm for point-cloud models based on data type conversion is
proposed. Firstly, a data type conversion rule-FtoI rule is designed. According to the FtoI rule,
float-point type point cloud is converted to integer type point cloud, then the integer type point-based
surface is split into many sized surface patches, the points of every patches construct a minimum
spanning tree, which is encoded in breadth first order. Besides we encode the difference between father
node and son node according to the minimum spanning tree, the difference is split into two parts, one is
sign, another is absolute value, which is encoded by arithmetic coding. Experiments show that this
compression algorithm has a nice compression speed and compression ratio without losing the quality
of point-cloud model.

Key words: 3D point cloud, lossy compression, float, the minimum spanning tree, arithmetic coding