The Research Of WMSN Video Image Reconstruction Algorithm Based On Block Compressed Sensing

As a distributed sensing network with the advantage of computing power, storagecapacity and communication capability, Wireless Multimedia Sensor Network (WMSN) canindependently complete monitoring task when deployed in an unattended environment, whichhas been widely used in many fields, such as transportation, military and industry. The maininformation carrier of WMSN is multimedia information, such as audio, video and staticimages, the traditional information processing technology is difficult to achieve the task oftransmission and storage because of its rich information and large amount of data.With the ability of accurately reconstructing the original signal with high probability by asmall number of measurements, a new way of compression named Compressed Sensing (CS)has been proposed, whose sampling rate only depends on the structure and content of signal,which breaks through the bandwidth constraints in the Nyquist sampling theorem. To a certainextent, it can alleviate the pressure of WMSN by using CS theory. In order to solve theproblem of computational complexity and storage space, Block Compressed Sensing (BCS)was put forward on the basis of CS, whose key part is information reconstruction. In theframework of BCS, this paper proposed a new WMSN video image reconstruction model bythoroughly analyzing the characteristic of video images. The main research work is asfollows:First of all, under the deep analysis of the texture features of single frame static image inWMSN, an adaptive sampling method based on the energy of image blocks was proposedaccording to the complexity of the image blocks. At the decoding end, an improved IterativeShrinkage thresholding (IST) reconstruction algorithm is proposed via introducing thepermutation weight matrix to construct the observation matrix of the original image.Consequently, on the basis of studying on redundant features in WMSN video image,combined with statistical information of WMSN video image in pixel domain and transformdomain, an adaptive sampling method based on the statistical information of image block waspresented. Besides, in line with Smoothed Projected Landweber (SPL) algorithm, animproved semi-iterative SPL algorithm was presented.Finally, in the light of redundant features and transformation features, Double TreeComplex Wavelet Transform (DT-CWT) method and the nonlocal similarity of the imageswere highlighted, and a new WMSN video image reconstruction model was established byintegrating the structural sparsity regularization and the nonlocal similarity regularization ofimages; on the basis of the model above, a modified Iterative Hard Thresholding (IHT) algorithm was proposed in accordance with the structure information in WMSN video images,such as edge, texture etc. Simulation experiment results show that the method can accuratelyreconstruct the original WMSN video images.