Area Efficient DST Architectures for HEVC

This work analyses the actual throughput of the Discrete Sine Transform (DST) stage in a realistic HEVC encoder, which executes the rate-distortion optimization algorithm to achieve high compression quality. Then, a low complexity DST factorization, where all the integer multiplications are substituted with add-and-shift operations, is exploited to design an efficient 1D-DST core. The proposed 1D-DST core is employed to derive two area efficient architectures, namely Folded and Full-parallel, for computing the 4×4 2D-DST in HEVC. Finally, the proposed 2D-DST architectures are synthesized on a 90-nm standard cell technology to support the actual target throughput required to encode 4K UHD @30fps video sequences, showing better area efficiency with respect to existing DST architectures for HEVC.