摘要
传统的DCT算法是直接采用DCT的定义进行变换,由于大量采用浮点运算,运算量大且精度差。针对DCT运算的复杂性,介绍了一种离散余弦变换(DCT)的快速算法,他采用基于尺度提升结构的矩阵分解法,将原先复杂的浮点运算转变为精度更高的整数运算,同时又将复杂的乘法计算分解为简单的移位和加法计算。因此在近乎无损地表示信息的同时,大幅度的降低了运算量。该方法非常适合于应用在实时性要求严格的场合,如视频压缩和目标跟踪,以及硬件的实时实现。最后用Verilog HDL语言给出了采用此算法的DCT核心计算模块的程序。
Traditional DCT algorithms, which transform directly by use of definition of DCT,adopt lots of float operation. Therefore,these algorithms have a mass of computation and low precision. Based on the complexity of DCT,an algorithm for DCT is presented. In this algorithm,a decomposition of matrices based on scale lifting structure is adopted. This algorithm not only convert the float computation to the higher precise integer computation,but also convert the multiplication to only shiftand - add operations. As a result,the information can be denoted loseless and the operational quantity is greatly reduced. This algorithm is very suitable for application to strict real time situation such as video compression,object track and implementation in hardware. In the end,this paper gives the Verilog HDL program of this algorithm.
出处
《现代电子技术》
2007年第8期98-100,共3页
Modern Electronics Technique
