Research On Distributed Arithmetic Coding For Binary Sources

With the rapid development of multimedia information,a compression scheme with high compression efficiency,high performance and stability is needed to process such a huge amount of information.Distributed arithmetic coding,as one of the implementations of distributed source coding,can not only meet the low-energy coding requirements,but also exhibit excellent performance in non-stationary and medium-short length data blocks.Therefore,it is widely applied to the field of wireless sensor networks,video coding,etc.Since the decoding performance of the distributed arithmetic coding usually depends on the use of the priori information of source and side information,how to use the priori information of source and side information is particularly important during the procedure of decoding.In addition,multimedia information sources such video sequence are usually with memory where they are relevant between symbols or symbol sequences.While the existing distributed arithmetic coding scheme mainly focuses on memoryless sources,not explore the compression effect of distributed arithmetic coding on memory sources.To overcome those problems,the main work of this thesis is as follows:1)In order to solve the problem that how to use the a priori information of source and side information during the process of decoding,this thesis proposes an modified decoding scheme where the decoding performance depends only on the correlation between source and side information.Existing distributed arithmetic coding schemes usually use the priori information at the decoding end.Different from the channel coding,distributed arithmetic coding use the prior probability to encode and store it in the codeword.So all possible decoding sequences should have the same priori information.In addition,according to the property of Markov chain among the codeword,source and side information,the proposed scheme simplifies the maximum a posteriori metric to a form where the decoding metric only relies on the correlation between source and side information.The experimental first studies the effect of the source termination length on decoding performance with different prior probabilities and different data block lengths of source,and then obtains the corresponding optimal termination length.Comparing to existing decoding schemes,the proposed scheme can achieve a better performance on different priori probabilities and different data block lengths of source.2)Due to existing schemes on memory sources are usually implemented by channel coding,this thesis proposes a scheme for compressing memory sources with distributed arithmetic coding.The proposed scheme firstly uses the correlation between source symbols and the property of independent and identical distribution between the source and side information.Then simplify the maximum a posteriori metric to the final decoding metric according to the property of Markov chain among the codeword,source and side information.Experimental results show that the decoding performance of the proposed scheme is better than some existing schemes on the first-order Markov source.In addition,we can find that the proposed scheme can obtain a good performance on the first-order Markov source with different overlapping factors.Finally,experiment results verify that the proposed scheme can achieve a good performance when applied to the second-order,third-order and fourth-order Markov sources with different data block lengths.This thesis first proposes a modified decoding metric of distributed arithmetic coding on memoryless sources where the decoding performance depends only on the correlation between source and side information.Then we also propose a scheme for compressing memory sources with distributed arithmetic coding.Experiment results show that those proposed scheme can achieve a better performance comparing to some existing schemes.