Research And Implementation Of AVS Encoder On Zedboard

With the advances in image processing technology, the design of videoencoder comes to mature. The paper realizes the AVS encoder based on AVSstandard and Zedboard. As a new developing board of Zynq-7000series,Zedboard not only integrates FPGA (Programmable Logic which can be calledPL), but also has a dual-coreARM (Processing System which can be called PS).For Zedboard has rich resources and a dual-platform architecture, the paperproposes a design based on it. PL Part is responsible for the hardwareimplementation of video capture and transform predictive. PS part is responsiblefor the realization of entropy coding through application program. Theycommunicate with each other throughAXI bus.PL part applies on-chip cloud architecture. It encapsulates functionalmodules into three atomic components, which are image capture atomiccomponent, luminance and chrominance prediction and transform atomiccomponents. They are deployed repeatedly by process engine to executemultiple operations in parallel. The architecture reduces the number of clocksand improves efficiency greatly. In addition, in order to reduce resourceconsumption, the paper proposes two optimization schemes. The first one iscircle queue, in which memory addresses are end to end. The storage scheme is quite necessary for the transmission of streaming data. The second one isrealizing the function of ram with VHDL language. It dramatically reduces ramconsumption and lays the foundation for parallel operation of on-chip cloudarchitecture.In part of PS, the entropy coding is realized in ARM processor. There aretwo algorithms in entropy coding, which are CABAC (Context-based AdaptiveBinary Arithmetic Coding) and CAVLC (Context-Adaptive Variable-LengthCoding). The paper analyses their algorithm principles and tests their encodingperformance in different QP and different video sequences. The results indicatethat their PSNR are almost at the same level, but the frame rate of CAVLC ismuch higher. In other words, CAVLC is more time-saving. Therefore, the paperapplies CAVLC algorithm as the entropy coding scheme to reduce encodingtime and make the video more smoothly.As described above, PL part realizes the image capture and prediction andtransform while PS part realizes entropy coding. Then how to transport thequantized coefficients of PL to PS is also the emphasis of the paper. The paperdetails how to build the channel between PL and PS to complete thetransmission of quantized coefficients. Then it verifies the performance of theAVS encoder by decoding the code streams of PS and displaying them on theclient. It shows that the design can realize real-time coding efficiently.