New texture coordinate representation exploiting texture patch formation

In order to improve the realization and rendering efficiency of a 3D mesh model, patch-base textures are widely used in 3D applications. However, the huge description size of texture patches - which is utilized for mapping texture patches onto 3D mesh - accounts for a substantial amount of storage and transmission resources. The well-known 3D mesh compression algorithms perform efficiently on connectivity and geometry data. In the mean time, less attention was given to texture coordinate compression. In this paper, we proposed an efficient representation method that can adaptively describe two common texture patch types: regular and non-regular patches. For textures with non-regular patch formation, we used a binary tree-like method in describing the connectivity between patches. For textures with regular patches, we proposed a new representation method by exploiting the regularity of texture patch shapes to describe a patch with one texture coordinate and a patch type (e.g., shape and orientation), not with three texture coordinates in the conventional representation. The proposed method is more compact than the conventional representation method (e.g., VRML), which leaves room for further compression of texture coordinate information. The experimental results showed that our algorithm produced a data reduction of 10 to 75 percent compared to the conventional representation method