图像算术编码特征及同步技术研究

Arithmetic coding technology plays a very important role for the performance of image compression. With the development of high precision space image data technology, there are still some issues for the improvement of image quality. Meanwhile arithmetic codec will de-synchronize when the single upset event or other noises cause coding circuit faults. Therefore the project will focus on performance and robust of arithmetic coding technology under space background. The detailed contents of our research consist of the following aspects. First, the combination of features in images and coding symbols is used for prediction of symbol's probabilities. We plan to find a good way for the mix of symbol probability with different features to improve the precision of symbol model. Second, we will study the characteristic of internal variables in arithmetic coding to obtain synchronization mode. We will propose a four dimensional state chart for clock synchronization. Third, the optimal strategy of the information redundancy allocation of arithmetic coding will be studied. We will propose a function with the constraint of distribution for different wavelet filter subbands. The solution is also described in the proposal. Finally, the project involves the prediction algorithm for asynchronization of arithmetic decoding. The proposal describes a maximum likelihood estimation based on the state path weight to improve the precise of prediction. And the error event queue theory model is used in the analysis of our structure. The detailed experimental plan is also presented. In a word, the project results can be used in various space image coding system to improve the ability of information acquisition and reduce the cost of external circuit reinforcement. It is meaningful for enhancement of space data transmission and other image coding scenarios.

算术编码是提高图像压缩性能的重要手段。当高精度空间图像应用时，图像编码质量与需求间仍然存在距离，同时由于粒子翻转或者噪声等引起编码电路错误会使得编解码失去同步,造成大量信息损失。为此，项目结合空间应用针对算术编码提高图像质量和鲁棒性展开工作，主要研究：1)编码特征与符号概率预测方法，研究多特征符号概率的混合，提高算术编码概率模型的准确度；2)编码内部变量同步模式，研究以四元组状态迁移时钟方式准确描述编码历经状态和刻画同步关系；3)信息冗余优化分配控制策略，研究依频带分布的受约束信息冗余分配极值函数和数值求解流程；4)算术解码失同步预测算法，研究按状态路径加权的极大似然估计方法，提高预测准确性，并给出错误翻转排队模型分析和验证实验。项目研究将会增强空间环境编码系统获取信息能力，对提升我国空间数传和通用图像编码水平具有实际意义。

项目组根据项目计划书，主要开展了图像算术编码研究工作，包括图像编码特征与算术编码性能研究、算术编解码内部编码区间、累积概率、码值等变量与输入符号位时钟同步关系研究、变换域系数信息分布与算术编码冗余优化分配研究和算术编解码失同步时解码位信息预测机制研究。项目研究成果已经应用于我国火星探测工程和某型卫星系统压缩项目中,取得了良好效果。项目提出了基于广度优先的SPIHT算术编码方案，多符号算术编码算法及解决方案，与现有星载压缩算法比较，取得了压缩性能约1dB的提升。项目组共发表论文20篇，其中SCI检索10篇，培养博士2名，硕士22名。