2M维矢量矩阵DCT整数变换及并行实现

2M-dimensional Vector Matrix DCT Integer Transform and Parallel Implementation

为了大幅降低多维数据在计算上所消耗的时间,提高计算速度,在2M维矢量矩阵DCT(2MVMDCT)整数变换理论的基础上,提出了将其进行并行处理的方法,增大时间优越性.首先介绍了2M维矢量整数变换核矩阵的推导过程;其次,将这种多维整数变换算法应用到多视角视频的压缩编码中,并与多维矢量离散余弦浮点变换进行能量集中性比较;最后,引入多核并行处理的思想,进一步提高处理速度.仿真结果表明,2M-VMDCT整数变换有着非常优越的能量集中性,将其并行实现能使运算效率大幅提高.

In order to improve the efficiency o f multi-dimensional data processing operations and reduce the computing time,a 2M-dimensional vector matrix DCT （ 2M -VM DCT ） integer transform parallel processing method is presented, according to the 2M-dimensional vector matrix DCT integer transform theory. First ly, the 2M-dimensional vector integer transform core matrix is introduced; Then,2M-dimensional vector matrix integer DCT transform is applied to the video compression coding, and compared with the energy concentration of the multi-dimensional discrete cosine floating-point transform; Final ly, by introducing the idea of multi-core parallel processing, the processing speed is improved. The simulation results show that 2M-dimensional vector matrix DCT integer transform has a better energy concentration and the parallel implementation improves its operation efficiency significantly.