基于分块压缩感知框架的无人机视频编解码研究

发布时间：2018-03-28 12:52

本文选题：分块压缩感知　切入点：视频压缩　出处：《西南交通大学》2017年硕士论文

【摘要】：无人机视频采集系统是以无线链路为通信信道,由无人机进行视频信息捕获,地面接收并分析的信息处理系统。无人机属于能量受限系统,通常采用低功耗、低性能的嵌入式设备作为视频获取设备。对于这类设备而言,低编码复杂度视频编码技术能够有效地提升系统性能。通过分析无人机视频采集系统与航拍视频流的特性,论文对现有的分块压缩感知图像编码框架进行了分析与研究,为了有效利用视频信号的时域相关性,论文在解码端引入运动补偿技术以提高视频流的整体压缩率。首先,为了满足无人机低实现复杂度的要求,在不影响图像重构质量的前提下,论文研究了基于哈达玛矩阵作为分块压缩感知系统的测量矩阵,以减少编码端的计算复杂度,提高编码速度;此外,论文研究了在编码端引入差分调制量化步骤,更有效地利用了视频信号帧内的空间相关性以及信息熵,提高率失真性能。其次,由于在实际场景中,解码端收到的信号往往带有噪声,为了能够高质量地恢复图像,论文在比较相关的降噪重构算法的基础上,验证了 D-AMP算法可以作为性能更为优异的重构算法。第三,考虑到视频流信号存在较强的时域相关性,虽然传统的运动补偿算法能够有效地利用时域冗余,但其复杂的计算将给编码端带来巨大的负担。论文借鉴分布式信源编码的思想,在解码端实现运动补偿技术。最后,考虑到由于无人机的高速运动,影响运动估计算法对图像块的匹配,造成视频质量的下降,论文提出了一种基于飞行状况的运动估计算法,该算法在传统运动估计算法之前加入预处理操作,根据当前帧的无人机运动状态信息设定搜索窗口的中心点位置,有效提高了运动估计算法的搜索精度。论文围绕无人机视频编解码的相关分析和研究对于后续进一步完善适用于无人机系统的视频编解码技术方案有参考和借鉴价值。
[Abstract]:UAV video acquisition system is an information processing system which uses wireless link as communication channel, video information is captured by UAV and received and analyzed on the ground. UAV is an energy limited system, which usually uses low power consumption. Low performance embedded devices are used as video acquisition devices. For such devices, low coding complexity video coding technology can effectively improve system performance. By analyzing the characteristics of UAV video capture system and aerial video stream, In this paper, the existing block compression perceptual image coding framework is analyzed and studied. In order to effectively utilize the temporal correlation of video signal, the motion compensation technology is introduced in the decoder to improve the overall compression ratio of video stream. In order to meet the requirement of low implementation complexity of UAV, without affecting the quality of image reconstruction, this paper studies the measurement matrix based on Hadamard matrix as a block compression sensing system to reduce the computational complexity of the coding end. In addition, the paper studies the introduction of differential modulation quantization in the coding end, which makes more effective use of the spatial correlation and information entropy in the video signal frame to improve the rate-distortion performance. Secondly, because of the actual scene, In order to recover the image with high quality, this paper verifies that the D-AMP algorithm can be used as a better reconstruction algorithm on the basis of the relative noise reduction algorithm. Considering that video stream signal has strong correlation in time domain, although the traditional motion compensation algorithm can make use of time domain redundancy effectively, its complicated calculation will bring huge burden to the coding end. This paper draws lessons from the idea of distributed source coding. Finally, considering that the high speed motion of UAV affects the matching of image blocks and the video quality, a motion estimation algorithm based on flight condition is proposed in this paper. In this algorithm, preprocessing operation is added before the traditional motion estimation algorithm, and the center of the search window is set according to the motion state information of the current frame of the UAV. This paper focuses on the analysis and research of UAV video coding and decoding, which has reference value for further improving the video coding and decoding technology for UAV system.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【相似文献】