Mesh segmentation schemes for error resilient coding of 3-D graphic models

Most existing coding techniques for three-dimensional (3-D) graphic models focus on coding efficiency. Due to irregular structure of the 3-D mesh and the use of variable length entropy codes, channel errors often propagate in the coded bitstream and severely distort the decoded model. By segmenting a 3-D graphic model (or its connected components) into small pieces, the impact of channel errors is confined to directly corrupted pieces rather than the whole mesh. The mesh segmentation schemes should be compatible with the underlying encoding techniques to achieve the low computational and coding overhead. In this research, we examine four mesh segmentation schemes, i.e., multiseed traversal, threshold traversal, morphing-based volume splitting, and content-based segmentation, and apply them to the context of error resilient mesh coding based on the constructive traversal coding technique. The advantages and shortcomings of each segmentation method are discussed. These schemes segment a mesh into pieces according to the target piece size that is determined according to the channel error rate.