【作者】 熊哲源；

【导师】 樊晓平；

【作者基本信息】 中南大学 ， 计算机应用技术， 2012， 博士

【摘要】 在资源受限的无线多媒体传感器网络中存在大量的图像数据需要处理和传输。为充分利用资源,需要设计一个能够综合考虑能量消耗、压缩率和图像质量之间平衡的图像编码方案。对无线多媒体传感器网络中图像编码技术面临的挑战和设计目标进行了讨论。对无线多媒体传感器网络图像编码现有解决方案和理论研究成果,分别从个体信源编码和分布式信源编码两方面进行分类探索。对个体信源编码存在的一些问题进行讨论,并指出了未来研究的方向。研究无线多媒体传感器网络中无线视频传感器节点的图像编码和图像传输的性能。对基于DCT和DWT的图像编码与传输的能量消耗和率失真进行建模分析。对适用于WMSN的图像质量评价指数进行研究,提出基于加权分块的峰值信噪比。根据所建立的能量消耗与率失真模型,对有限的能量和传输带宽进行优化分配,以使得能量消耗最小化。仿真实验结果表明,所提出的方法在保证应用所需图像质量的情况下,有效地降低了无线传感器节点的数据通信量和计算过程的能耗。研究无线多媒体传感器网络中视频监控图像序列的压缩,提出一种基于变化检测和改进JPEG算法的低复杂度图像编码方案。通过变化检测算法定位监控图像中的运动目标即感兴趣区域,仅传输兴趣区域以减少数据传输量,适应无线传感器节点存储转发能力有限的特点；改进JPEG算法中的DCT和量子化过程以降低计算复杂度,适应无线传感器节点计算能力有限的特点。算法复杂度分析和仿真实验结果表明,所提出的方法在保证应用所需图像质量的情况,有效地降低了无线传感器节点的数据通信量和计算过程能耗。为充分利用有限的资源产生高分辨率、宽视角图像,考虑到相邻无线视频节点之间的信息冗余性,提出一种适用于无线多媒体传感器网络的图像拼接算法。使用分块搜索算法进行图像配准以降低能耗,改进绝对差值和算法以提高图像配准的精度,并使用渐进渐出的加权平均算法对图像进行缝合,图像拼接之后簇头与基站间的通信量减少,可以有效降低网络负载。仿真实验结果表明,所提出的算法在保证一定图像配准精度和图像质量的情况下,计算复杂度较低,可以有效节约能量。针对图像质量要求较高的应用场景,提出了一种能量有效的JPEG2000图像编码算法,根据网络条件和图像质量的限制,使用查找表来选择适当的量子化层级和小波变换层级以减少能量消耗。并采用给予优先级的图像传输方案,根据节点剩余能量和数据优先级来决定转发或丢弃。仿真实验结果表明,所提出的方法能够在保证所要求图像质量的情况下,有效地降低无线传感器节点的计算和通信能耗。更多还原

【Abstract】 There are a large number of video data have to be processed and transmitted in resource-constrained wireless multimedia sensor networks (WMSN). One possible way to achieve the maximum utilization of those resources is to apply an adaptive video coding scheme, which must consider the trade-off among energy consumption, compression ratio and image quality. The state-of-the-art in video coding techniques, the major research challenges and the objectives of video coding for WMSN are discussed. Existing solutions and theoretical explorations of video coding for WMSN are investigated and classified, including single source coding and distributed source coding. Finally, fundamental issues of single source coding are discussed, and future research trends in this area are outlined.The image communication behavior of a wireless video sensor and its performance under resource constrained wireless multimedia sensor networks is studied. Energy consumption and rate distortion model of DCT and DWT based image compression and transmission is developed. Based on energy consumption and rate distortion model proposed, resource allocation is optimized with limited energy and bandwidth. Encoding coefficients are adapted according to the activity of monitoring scene, and then the energy consumption will be minimized. Simulations results are conducted to show the performance of our work. The proposed scheme dramatically reduces image compression and transmission energy consumption under expected image distortion and transmission rate.With resource-constrained wireless multimedia sensor networks, image coding and transmission must respect the trade-off among energy consumption, compression ratio and image quality. The problem of compression of video-surveillance image sequences collected by a wireless multimedia sensor network is studied. To reduce the computation complexity, a low-complexity image compression scheme based on change detection and adapted JPEG is proposed. Change detection is used to locate the region of interest and cut down the data for transmission, fast DCT is used to reduce the computation complexity. Such an image compression scheme provides a graceful trade-off between the reconstructed images quality and the sensor nodes’lifetime. Computation complexity analysis and simulations on the parking lot images are conducted to show the performance of our work.For sufficient utilization of constraint resource on generating high-resolution images with wide field of view, an algorithm takes fully account the redundant visual information among multiple video sensors is proposed. Image block searching algorithm is applied for image registration to decrease the energy consumption, sum of absolute difference algorithm is adapted to improve the accuracy of image registration, and weighted mean algorithm is applied for image stitching. Data volumes of transmission are decreased after image mosaicking, which can efficiently reduce the network loading. Simulation results demonstrate that the computation complexity of proposed algorithm is lower than other image mosaicking algorithm, under certain image registration accuracy and image quality.An energy efficient JPEG2000image compression scheme is adapted for wireless multimedia sensor networks. The algorithm determines the parameter of quantization and the wavelet image compression transform level to meet the image quality requirement and channel condition. Semi-reliable transmission enables packet forward or discarding by intermediate nodes according to residual energy and data priority. Simulation results show that proposed image compression scheme provides a graceful trade-off between the reconstructed images quality and the total energy consumption of sensor nodes.更多还原