发明名称 Compressing data representing computer animated hair
摘要 Data representing animated hair in a computer generated imagery (CGI) scene may be compressed by treating hair data as arrays of parameters. Hair data parameters may include control vertices, hair color, hair radius, and the like. A principal component analysis (PCA) may be performed on the arrays of hair data. PCA may yield new basis vectors, varying in length, with the largest basis vector corresponding to a new dimension with the largest variance in hair data. The hair data may be quantized based on the varying lengths of new basis vectors. The number of bits allocated for quantizing each new dimension corresponding to each new basis vector may be determined based on the relative lengths of new basis vectors, with more bits allocated to dimensions corresponding to longer basis vectors. The quantized hair data may be bit-packed and then compressed using lossless entropy encoding.
申请公布号 US9064345(B2) 申请公布日期 2015.06.23
申请号 US201313830943 申请日期 2013.03.14
申请人 DreamWorks Animation LLC 发明人 Matthews Mark Jeffrey
分类号 G06K9/00;G06T13/40;G06T9/00 主分类号 G06K9/00
代理机构 Morrison & Foerster LLP 代理人 Morrison & Foerster LLP
主权项 1. A computer-implemented method for compressing data representing animated hair in a computer generated imagery (CGI) scene, the method comprising: receiving, by one or more processors, data representing animated hair, wherein the data comprises a plurality of arrays of hair data parameters; determining, by the one or more processors, a set of new basis vectors for the plurality of arrays of hair data parameters using principal component analysis (PCA), wherein the number of new basis vectors corresponds to the number of hair data parameters in an array; determining, by the one or more processors, an allocation of bits to dimensions corresponding to the set of new basis vectors based on the relative lengths of the new basis vectors, wherein a new basis vector with the greatest length is allocated more bits than a new basis vector that is shorter; quantizing, by the one or more processors, the hair data in the dimensions corresponding to the set of new basis vectors based on the allocation of bits; bit-packing, by the one or more processors, the quantized hair data; compressing, by the one or more processors, the bit-packed quantized hair data, wherein the compression is lossless; and storing the compressed hair data.
地址 Glendale CA US