Research On Image Coding Compression And Transmission In Wireless Multimedia Network

As a new type of regional observation technology,wireless multimedia sensor networks(WMSNs)have won highly attention of domestic and international researchers.In a densely deployed WMSN,the visual information observed by camera nodes with overlapped field of views generally exists redundancy and certain space correlation.Subject to the limits of energy resources and communication bandwidth,WMSNs should be designed to deliver multimedia content with a certain level of quality-of-service.This thesis focus on the research of image compression coding,image transmission and image stitching on the context of Wireless Multimedia Sensor Networks(WMSN).This thesis introduces software and hardware parts of the structure of the network nodes system in WMSN which are based on the status domestically and internationally.We analyze the hardware resources and function of DM6437 and establish embedded image compression and transmission platform based on DM6437.Synthetic and real experiment demonstrate the high speed of image acquisition and transmission using the Zigbee.After deep discussion of integrated development environment,we investigate the image acquisition and compression theory with respect to the software.Considering the limited network bandwidth of WMSN,this thesis proposes an improved DCT transformation method where quantization table is adapted according to the specific image compression in JPEG compression.Then,the experiment utilizes peak signal to noise ratio(PNSR)parameter index decides the ultimate and best parameter of JPEG compression.Real experiment demonstrated the high compression ratio of the proposed method with compression ratio 2.97 times the traditional method when peak signal to noise ratio(PNSR)is less than 33.In order to improve the data processing efficiency of the camera nodes,in this thesis,optimize the code and the real-time of the embedded platform.Then,the execution efficiency of DSP is improved greatly.According to actual circumstance that there exists overlapping area among the camera nodes.This thesis put forward the theory which combines with the model of nodes perception and the deployment of the cameras.Then calculate the spatial correlation between the nodes to describe the relevant characteristics of visual information in the interest region.This thesis put forward the theory which based on entropy analysis framework.Then use the theory to measure the vision information which provided by cameras in WMSN.At the same time,the camera nodes which were selected according to the relevant function and entropy framework.Then,image stitching can be finished combined with spatial geometrical-restriction relation.The last part is a summary of the research in this thesis,with a prospect of the further work.