论文部分内容阅读
提出一种基于查表法的二维8×8离散余弦逆变换(2D 8×8 IDCT)的快速算法,其查找表LUT(Look-UpTable)结构的设计是基于二维8×8 DCT的基本图像.利用两种技术减小查找表长度:①利用基本图像的对称特性;②通过对离散余弦正变换(DCT)和量化过程的分析,推导出每个量化后DCT系数的取值范围.使得查找表只有10.9746K项数据,若量化矩阵具有对称性q(u,v)=q(v,u),LUT的长度还可减少近半.新算法利用查表法消除IDCT中乘法运算,并利用图像数据的特点和基本图像的对称特性大大减少加法次数,提高了计算速度.以多幅标准图像为样本数据进行实验,结果表明:新算法实现2D 8×8 IDCT运算平均只需加法182次.与当前运算量最小的Feig快速算法做比较,新算法避免了乘法,所需加法次数也降低了约15%.
A fast algorithm of 2 × 8 × 8 discrete cosine transform (2D 8 × 8 IDCT) based on look-up table is proposed. The design of look-up table (LUT) is based on the basic of 2 × 8 × 8 DCT Two techniques are used to reduce the length of the look-up table: (1) the symmetry of the basic image is utilized; (2) the range of values of each quantized DCT coefficient is deduced by analyzing the discrete cosine transform (DCT) and the quantization process The lookup table has only 10.9746K data, and the length of the LUT can be reduced by nearly half if the quantization matrix has symmetry q (u, v) = q (v, u) .The new algorithm eliminates the multiplication of IDCT The use of the characteristics of image data and the symmetry of the basic image can greatly reduce the number of additions and increase the computing speed.Experimental results using multiple standard images as sample data show that the new algorithm only needs 182 additions Compared to the current Feig fast algorithm, which minimizes the amount of computation, the new algorithm avoids multiplication and reduces the number of additions by about 15%.