UAV Video Coding Scheme Based On Block Compression Sensing Framework

UAV video acquisition is an information processing system,wherein the video information captured by UAV will be delivered to the ground information processing unit via the wireless link.The UAV is a typical energy constrained system,wherein embedded devices with low power consumption and low realization complexity is usually assumed.Due to the wireless transmission,low encoding complexity video acquisition effectively guarantee the video codec applications thus becomes highly desirable.In this thesis,a survey is presented to analyze the features of different UAV video acquisition system and to address the characteristics of the aerial video stream.And it is shown that the block-based compressed sensing image encoding framework provides a promissing technical solution to the UAV viedo acquisition application.In order to effectively suppress the temporal correlation of the video stream,the motion compensation technique was proposed to improve the overall video compression efficiency.And the work in this thesis can be summarized as follows.Firstly,in order to meet the low realization complexity requirement of the UAV system,Hadamard matrix block based compressed sensing measurement matrix is recommended to improve the encoding efficiency and to reduce the calculation complexity.Secondly,the differential modulation quantization step in encoding is proposed in order to further suppress the spatial correlation of intra-video signal redundancy and to exploit the information entropy to improve rate distortion performance.Thirdly,in order to effectively cope with the transmission imperfections owing to the interference,after a comprehensive comparison of the noise reduction signal reconstruction algorithm,it is verified that the D-AMP algorithm is an efficient noise reduction algorithm with excellent reconstruction performance.It is known that there exists noticeable temporal correlation in the video stream.although the traditional motion compensation algorithm can be employed to exploit the time domain redundancy,but the involved complexity calculation will impose a serious burden for practical implementation.By following the distributed source encoding idea,the implementation of motion compensation at the decoder side is investigated.Finally,considering the high velocity of the UAV and its degradations of the videa quality in the image block matching based motion estimation,the thesis puts forward a preprocessing step to introduce flight status based motion estimation in the motion estimation.It is shown that,by introducing the pre-processing stage before the UAV motion state information set search window in current frame,the accuracy of the motion estimation algorithm can be improved effectively.This analysis in this thesis may provide useful reference for the further research of the UAV video coding and decoding framework.