The efficient implementation of 3-D transforms is a challenging task due to the computation complexity, memory and area requirements of such transforms. One important 3-D transform is the 3-D Discrete Cosine Transform (3-D DCT) used in many image and video processing systems. In this paper, two new pipeline architectures for the 3-D DCT computation using the 3-D DCT Vector-Radix algorithm (3-D DCT VR) are presented. These architectures are scalable and parameterisable with regards to different wordlengths and pipelining levels. Their arithmetic component requirements are reduced to the order of in contrast with for 3-D DCT architectures in the literature, while at the same time they can keep similar or better area-time complexity.