Research And Implementation On Architecture Of Remote Sensing Image Compression Encoder

In the current image application fields, remote sensing image which is a wide rangeand high reliable way to get information plays an increasingly important role in thenational economy. However, with the advances in image sensor production technology,the rapidly increased image information causes issues for processing system. Thus,aiming at reducing data storage costs, easing the communication network bandwidthpressure and breaking the processing frequency limit, it is desirable to study andimplement a high performance image compression encoder. In this paper, we make thefollowing research.Firstly, structure and principle of the remote sensing image compression encoder isintroduced. Based on the H.264/AVC protocol, it uses the optimal KLT transform andestimates bit rates by CAVLC, resulting in better performance and higher difficulthardware implementation than existing ones.Secondly, the three pipelines on architecture of the remote sensing imagecompression encoder have been proposed by weighing the design constraints. The firstlevel pipeline stores and assigns remote sensing image. The second one implementsdirection predicting and rate-distortion comparison of I4/I8/I16with parallelism and thelast one achieves the CABAC using two pipelines. To optimize the structure andincrease data throughput, I4/I8adopts nine-pipelines and I16uses serial way.Thirdly, the direction predicting circuits and reference pixel caches circuits in thehardware architecture are realized. With the hierarchical storage structures, the differentblocks reference pixels exchange quickly and avoid conflicts. By using redundantmicro-code to control the reusable direction predicting circuits in I4/I8, the consumptionof its resources greatly reduces compared with the case that all direction predicting areparallel circuits in the experiment.Finally, this paper describes a real-time processing system of remote sensing imagecompression encoder. In the system, the compression encoder circuit communicateswith man-machine interface via a USB cable. In addition,42remote sensing images ofdifferent features are used to test the system, and the statistical results show that itsaverage PSNR is0.5dB higher than JPEG2000.