In this paper, we propose an array-based architecture for DCT computation with high scalability. The scalable architecture can perform DCT computations for 15 different zones and 8 different precisions to achieve quality scalability for DCT coefficients. Due to the quantization process in video coding, the quality can still be retained for larger quantization parameter. We show the detailed comparisons between the quality scalability and the tradeoff factors, i.e., throughput, hardware resources, clock frequencies, and power consumptions.