| Wireless sensor networks (WSNs) are very important wireless computer communicationnetworks emerged in recent years. Because of the great potential for applications, WSNs haveharvested numerous research achievement and been widely adopted in a lot of fields. Along with theprogress of applications in WSNs, from simply measuring scalar physical phenomena liketemperature, humidity, light intensity, vibration, etc., WSNs gradually extended to use multimediainformation, such as images, audio, videos for finer granularity and accurate environmentalsurveillance activities. This extended sensor network also is a kind of WSNs with sensor deviceequipped with multimedia tools in essence, also called Wireless Multimedia Sensor Networks(WMSNs). Compare with conventional WSNs, this extensional WSNs can sense rich multimediainformation, like, audio, videos and images, would span a wide spectrum in military, industrial,commercial and environmental monitoring. One important research aspect of WMSNs ismultimedia information processing, whose objective is to achieve the high efficiently images andvideos multimedia information compression coding and transmission on the constraint of singlenode power supply energy, communication capacity, compute and memory in WMSNs. Theresearch objective of this dissertation is to further investigate the mechanism of images and videoscompression and transmission in WSNs. By the studies of the sensing model of multimedia node,the models of image redundant and node correlation and the coverage optimization and enhancingmethod for the whole network, to propose the novelty, distributed and node cooperative images andvideos compression and transmission techniques. Based on these, we try to achieve a WMSNsmultimedia information compression and transmission mechanism with less network traffic, longernetwork lifetime and better quality of transmission. The main contents and contributions of thisdissertation are as follows:(1) Node scheduling scheme for spatial-temporal coverage optimization. In sensor networks,coverage reflects the sensing range form physical world to sensor nodes and is the basis of the dataacquisition and processing in sensor networks. In sensor networks, based on the novel3D sensingmodel, two kind of network elementary regions generation methods, grid-based and position-basedare proposed. Then the spatial-temporal optimization scheduling algorithm is designed. A set ofsimulation experiments are put forward to evaluate the performance of the proposed grid-based andposition-based elementary regions generation methods. Further simulations show that the proposedspatial-temporal optimization scheduling scheme effectively enhances the performance ofspatial-temporal coverage optimization.(2) Multi-node cooperative image compression scheme based on Singular ValueDecomposition (SVD). For the issue of image compression in sensor networks, an adaptive blocking image compression algorithm based on SVD is studied firstly．Then, starting from thenode collaboration features, according to the roles division, camera nodes and common nodes arecooperated to accomplish the workload of image acquisition, compression and transmission.Simulations show proposed scheme can effectively compress and transmit images, balance theenergy consumption of network and prolong the life of the whole network.(3) Camera nodes correlation model based on3D sensing. Based on the analysis of novel3Dsensing model, further investigate the node cooperative model in (2). First, an asymmetricalrelationship correlation model is proposed. Then, two kinds of cluster structure, camera sensornodes cluster, and common sensor nodes cluster are established to cooperate on image processingand transmission tasks. Simulations show that the proposed network topology and image fusion andtransmission scheme released the pressure of camera node greatly and reduce the network energyconsumption of communication of the whole network efficiently.(4) A novel side information generation framework for multiview distributed video coding.Distributed video coding is adopted to process the multiview videos in sensor networks. motionintense regions (MIRs) and non-motion intense regions (NMIRs) based on sum of absolutedifference (SAD) criteria are distinguished. For the MIR, the side information (SI) is generated byfusion temporal SI and interview spatial SI at the pixel level. But for the NMIR, the temporal SI isdirectly use as the ultimate SI. Experimental results show that the proposed fusion SI approach canbetter performance when compares with only temporal SI used. |
PDF Full Text Request