Research On The Design Of Fully Hardwired CABAC Encoder For Remote Sensing Images

With the development of remote sensing, it is becoming much more difficult tocompress and store remote sensing image with much higher spatial resolution, temporalresolution or spectral resolution. It is critical to develop the compression technology forremote sensing image. My work focuses on high performance circuit design of fullyhardwired entropy encoder for remote sensing image, mainly to achieve Context-basedAdaptive Binary Arithmetic Coding (CABAC). CABAC is the entropy coding tooladopted in Main and High profiles of H.264/AVC video coding standard. CABACprovides significantly higher compression ratio than Baseline profile entropy coderCAVLC, with significantly increased computational complexity. Due to the sequentialcoding nature of CABAC with strong data dependency and frequent memory access, itrequires much computing time to implement CABAC encoding by softwareoptimization. Therefore, hardware implementation of CABAC encoding is necessary inhigh-rate image or video encoding.The main contributions in this thesis are as follows. First, we design the multilayercontrolling units, which ensure that every function unit works properly, and propose aparallel processing architecture based on buffer, which effectively solve the problem ofmismatching data rate between read and write. Second, we optimize the number ofcontext for remote sensing image and adopt tightly coupled method for transmission ofdata to reduce RAM read/write frequency. Third, we provide a4-stage pipelinearithmetic encoder with forwarding logic for context update, to solve data collision insingle cycle pipeline and to speed up encoding time. Finally, we design reused circuit toimplement three encoding model to reduce hardware resource and the size of thecircuits.A maximum processing speed of120MHz has been obtained by simulating withISE Design Suite12.2on Virtex6XC6VLX240T Field-Programmable Gate Arrays(FPGA) chip. Finally the CABAC encoder acquires high performance cooperating withother compression system on XC6VLX240T chip and meets the requirement for remotesensing image compression. The experimental results show that the CABAC encodercan process one bin per cycle and avhieves15%bit-rate savings as compared to KTA2.2with CAVLC.