A novel architecture for fast 2D IDCT decoders, with reduced number of multiplications

This paper presents a new architecture for the fast computation of the 8×8 two dimensional (2D) Inverse Discrete Cosine Transform (IDCT). The key idea of the proposed method is the permanent storage of the Basis Matrices of the 8×8 2D Discrete Cosine Transform (DCT). By exploiting the sparseness property of the 2D DCT coefficient matrix, the computational time decreases as the number of nonzero coefficients decreases. Moreover, by exploiting the symmetry properties inherent in the Basis Matrices, the introduced method reduces the number of multipliers as well as the number of required multiplications. The proposed structure computes all 64 pixel luminance values of an 8×8 block simultaneously. The efficiency of the proposed architecture is validated through simulations.