Abstract
The HEVC video coding standard supports different transform sizes ranging from 4-point to 32-point. In fact, multiple transform sizes improve coding efficiency, but increase as well the computational complexity. Hardware decoders apply different techniques to satisfy real-time requirements. This paper describes a novel design methodology of a unified 2D inverse core transform IICT. The hardware architecture is based on a 1D-IICT block and a transpose buffer FIFO memory used to store the intermediate values of 1D transform. All this process is controlled in such a way to reduce the hardware and memory resources. To support the different transform sizes, matrix multiplications are simplified based on transform blocks decomposition into fixed-size sub-blocks in previous works. The architecture was developed for an FPGA device. Synthesis results on Startix III FPGA device show that the proposed design, operating at 266 MHz, is sufficient to decode high resolution videos using only 10% of total pins and about 33% of the hardware resources offered.