Homogeneous Description for Heterogeneous Cross-Media Resources

Images, polygonal models and point clouds are three kinds of heterogeneous cross-media resources, but they can stand for the same object. We introduce two kinds of homogeneous descriptions for them. First, we find a relation between surface moment invariants and shape distribution since they both use invariant geometric primitives. This relation can provide a way to approximate surface moment invariants for point clouds. Hence, we can get the same shape description for both point clouds and polygonal models. Second, we project each 3D model in many directions and get large numbers of 2D images. Then, polygonal models and images are converted to the same data format for comparison. Homogeneous shape description makes it possible to carry out heterogeneous cross-media retrieval in pervasive environment.