《High-efficiency pipeline design of binary arithmetic encoder》.pdf

SCIENCE CH INA InformationSciences · RESEARCH PAPER · September2014，Vo1．57 092108：1-092108：8 doi：10．1007／s11432—013—4942—2 H igh—efficiency pipelinedesign ofbinary ● ■ 1 ● ● 1 arlt11m etlC enCoCLer SONG Rui CUIHongFei，LIYunSong＆W U ChengKe StateKeyLaboratoryofIntegratedServiceNetworks，XidianUniversity，Xi’an710071，China ReceivedSeptember14，2013；acceptedNovember21，2013；publishedonlineJanuary24，2014 A bstract This paper focuses on the pipeline design ofcontext—based adaptive binary arithmetic coding (CABAC)．CABACisawell—knownbottleneckinverylargescaleintegrationcircuitdesignofH．264／AVC encoder．Despiteitshighperformance，thetightfeedback loopsofCABAC makeparallelizationdifficult ． Most researchersareconcerned aboutmulti—bin processing regardlessofpipelinedesign．However withoutpipeline， ， theoverallperformancebecomessignificantly limited．In thispaper，thecriticalpath forthehardwareimple- mentationofbinaryarithmeticencoder(BAE)wasanalyzedindetail．Webreakdownthecomputingsteps to thebestextent，and rearrangesuch stepstotheappropriatepipelineto achieveabalanced latency ateach stage． M oreover，a new BAE architecturewith a five—stagepipeline and one bin percycle isproposed the ， latency