부산대학교 도서관

This paper presents a parallel and scalable solution for adaptive deblocking filtering in H.264/AVC. While traditionally in deblocking filtering, the edges in a macroblock are processed in a sequential order, this paper demonstrates how algorithm modifications can be used to enable processing multiple consecutive edges at the same time. The proposed method increases the throughput in proportion to the number of edges that are being processed simultaneously without affecting the PSNR and bit-rate. Details of the method to process 2 consecutive edges in parallel as well as extensions to process 4 and 8 consecutive edges, are provided. A dedicated hardware architecture to process 2 edges is presented along with synthesis results. The architecture achieves a 2× increase in throughput at the expense of a 2.2× increase in area and a 1.23× increase in power.