高效能H.264 MVC去區塊效應濾波器架構設計
郭其昌
Abstract
郭其昌
Abstract
In this paper, a double-filter design of deblocking filter is proposed to support Stereo High Profile of H.264 Multiview Video Coding (MVC) extension. Without using SRAM for local memory, the proposed design adopts register arrays with a novel data exchange mechanism to efficiently reuse the intermediate data while filtering picture coded in macroblock adaptive frame field (MBAFF) format. Cooperated with the proposed parallel filtering orders for double-filter design, the proposed filter takes 147 cycles per macroblock in average. It is synthesized under 0.13m CMOS technology and it can be operated at 250 MHz. At the cost of 55.2K gates with 0 byte SRAM, the proposed design achieves the throughput of 1,700,680 macroblock/sec. Thus, it meets the requirement of 1920x1080 at 120 fps for multiview video applications.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In this paper, a double-filter design of deblocking filter is proposed to support Stereo High Profile of H.264 Multiview Video Coding (MVC) extension. Without using SRAM for local memory, the proposed design adopts register arrays with a novel data exchange mechanism to efficiently reuse the intermediate data while filtering picture coded in macroblock adaptive frame field (MBAFF) format. Cooperated with the proposed parallel filtering orders for double-filter design, the proposed filter takes 147 cycles per macroblock in average. It is synthesized under 0.13m CMOS technology and it can be operated at 250 MHz. At the cost of 55.2K gates with 0 byte SRAM, the proposed design achieves the throughput of 1,700,680 macroblock/sec. Thus, it meets the requirement of 1920x1080 at 120 fps for multiview video applications.
Key concepts: Macroblock, Deblocking filter, Computer science, Computer hardware, Static random-access memory, Filter (signal processing), Byte, Coding (social sciences)