A video rate 16*16 discrete cosine transform IC

Abstract

The first fully operational implementation of a complete 16x16 Discrete Cosine Transform (DCT) on a single chip with the speed and the resolution necessary for video images is described. Using a pipelined concurrent architecture, the DCT IC was designed for real-time processing of 14.3 h4Hz sampled video data. Distributed arithmetic combined with bit-serial and bit-parallel structures is used to implement the required vector inner products concurrently. The chip accepts 9-bit inputs and produces 14-bit DCT coefficients. The circuit has been laid out using the symbolic design tool MULGA. The chip contains approximately 73,000 transistors and was fabricated using a double metal N-well 2pm CMOS technology.