Vertex data compression through vector quantization

Abstract

Rendering geometrically detailed 3D models requires the transfer and processing of large amounts of triangle and vertex geometry data. Compressing the geometry bitstream can reduce bandwidth requirements and alleviate transmission bottlenecks. In this paper, we show vector quantization to be an effective compression technique for triangle mesh vertex data. We present predictive vector quantization methods using unstructured codebooks as well as a product code pyramid vector quantizer. The technique is compatible with most existing mesh connectivity encoding schemes and does not require the use of entropy coding. In addition to compression, our vector quantization scheme can be used for complexity reduction by accelerating the computation of linear vertex transformations. Consequently, an encoded set of vertices can be both decoded and transformed in approximately 60 percent of the time required by a conventional method without compression.

Dynamically restricted codebook-based vector quantisation scheme for mesh geometry compression

“…The coordinates are then uniformly quantized with entropy coding using a modified Huffman code. Since then, many variations of Deering's scheme have also been proposed [7,14,15,[19][20][21][22]. Karni and Gotsman [23] demonstrated the relevance of applying quantisation in the space denoted by spectral coefficients.…”

“…Since vector quantisation (VQ) techniques exhibits many superiorities over scalar quantisation methods, VQ schemes have also been proposed for 3D mesh compression in [21,22] since 2000. In Lee and Ko's work [21], the Cartesian coordinates of a vertex are transformed into a model space vector using the three previous vertex positions and then quantised.…”

“…In Lee and Ko's work [21], the Cartesian coordinates of a vertex are transformed into a model space vector using the three previous vertex positions and then quantised. In Chou and Meng's work [22], several VQ techniques including the open loop VQ, the product code pyramid VQ and the asymptotic closed loop VQ are applied for residual vector quantisation.…”

“…In order to exploit the correlation between vertices, it is necessary to use a vector quantiser with memory. We employ an approach known as predictive vector quantisation (PVQ) [22]. The index identifying this residual vector in PVQ is then stored or transmitted to the decoder Fig.…”

Geometry-aware bases for shape approximation

“…The basis is derived from the spectral decomposition of the Laplacian matrix associated with the mesh connectivity [10]. Chou and Meng [8] encode the geometry of the mesh using vector quantization of the displacement coordinates. Based on an analysis of the spectral basis of the Laplacian, Sorkine et al [9] introduce a method where the quantization is applied to the geometry vector transformed by the Laplacian operator.…”