无线传感器网络中的多媒体视频压缩算法分析-analysis of multimedia video compression algorithm in wireless sensor networks.docxVIP

南邮电大学硕士研究生学位论文AB 南邮电大学硕士研究生学位论文 ABSTRACT II II ABSTRACT As a new and emerging type of sensor networks, Wireless Multimedia Sensor Networks have applied more and more for completely monitoring the target enviroment.Different from the traditional Wireless Sensors Networks which usually measure scalar physical phenomena, Wireless Multimedia Sensor Networks focus on the capture and processing of large multimedia information like audio and video stream. In the Video application, the video compression can reduce the data information of Wireless Sensors Networks. Nowadays, many typical Video Compression Coding Standards (like MPEG-X, H.26X) adopt motion compensation on the Encoder End. So the complexity of Encoder is 5 to 10 times then the Decoder that cant met the demand of Wireless Multimedia Sensor Networks which not only have a limited network resource but also the low complexity and low power consumption node. Since the Distributed Video Coding adopt Intraframe encode + Interframe decode and excavate the dependence of video sequence on the Decoder, so the complexity of Encoder is lower and make the Encoder more brief. The distributed video coding is very applicable to solve the video coding issues on the Wireless Multimedia Sensor Networks. This thesis is based on the Distributed Video Coding. Aim at the issues that traditional Distributed Video Coding cant decode the intense motion and boundary regions exactly, this thesis proposed an algorithm which based on gradient field. The algorithm utilize surface fitting to produce the image gradient field and extracts such areas through the criteria of region of interest, then decode such areas by Expectation Maximization iteration algorithm and decode other areas by LDPC. Since the quantification and Discrete Cosine Transformation on the encoder will bring up the block-effect and the algorithm apply Deblocking filtering on the decoded picture. Simulation results show that: the algorithm not only ca