The Research Of WMSN Video Image Reconstruction Algorithm Based On CS

Wireless multimedia sensor network (WMSN) is emerging on the basis of wirelesssensor network (WSN), so both of them are independent networks with self-organization.Image and video which become the main information carriers are imported in WMSN,therefore, WMSN is widely used in military and industrial areas. However, according to theNyquist sampling theorem which is used to sample the image there will produce amounts ofdata. As a result, it requires very high hardware devices which implement the tasks oftransmission and storage. What’s more, it is a challenge for WMSN with limited energysensor node.Recently, there is emerging a new compressed undersampling method called compressedsensing (CS) which states that sparse signal collected under very low sampling rate can bereconstructed highly accurately. Because CS is inherently sensing signal directly inaccordance with content and structure in signal, so it relieves the pressure on WMSN throughCS technology which can reduce the amount of data representing signal. Reconstructionalgorithm that is used to recover the signal in CS is the key to the video image processingbased on CS. In this context above, the paper researches the video image CS reconstructionalgorithm on the basis of exploring the priori information of the image in WMSN. On theprevious work an effective reconstructed method was proposed to improve the quality of theimage recovered, which promotes the application for WMSN. The main content andinnovation of the paper are summarized below:The paper introduces the CS theory and related concepts in the emerging field. On thisbasis, the CS reconstruction algorithm is studied. At the same time, the experiments wereimplemented based on the two kind methods, which laid the foundation for later work. Anon-local regularized WMSN video image iterative sparse reconstruction was proposed on thebasis of researching the CS reconstruction algorithm. Because priori information is thereconstruction condition of the application of CS, the rich structure feature of video image isstudied. As a result, a new video image reconstruction model is formed based on non-localregularization which is constructed by taking image self-similar into account. An improvedalgorithm is proposed based on iterative shrinkage thresholding algorithm to resolve themodel. In the reconstruction process, the geometry feature priori of the image is promotedbesides sparsity priori. In experiments, compared with reconstruction image by other model,the quality of image by proposed algorithm in this paper has better visual effect whichpreserves the fine information of original image.Meanwhile, a scheme of WMSN video image coding/decoding is designed for thespecific application scenario such as traffic surveillance. By the utility of the CS theory, the scheme is divided based on both intra-frame and inter-frame which largely reduces theamount of transmission data resulting in saving the network energy consumption. The keyframe image is coded in intra-frame based CS. The non-key frame images are coded jointly ininter-frame based on the frame difference technique and sparse subspace clustering thought aswell. The video images are decoded based on non-local regularized sparse iterativereconstruction algorithm. Simulation experiments show that the scheme can reduce theencoding complexity and reconstruct the high quality images simultaneously, so as to ensurethe reliability of WMSN traffic surveillance.