Image Fusion Method Based On Multi-morphological Sparse Representation In WMSN

As an important technology to perceive the objective world，WMSN has been widelyused in network monitoring and other fields. However, due to the complex monitoring scenein WMSN, it is vulnerable to lose important information for WMSN by only using singletype image sensors to capture images. Moreover, tremendous huge amounts of image dataneed to be transferred in WMSN，the life-time of network will be dramatically reduced. Thus,it is imperative to decrease the data transmission in order to prolong the life-cycle of network.On the basis of CS and image fusion theory, a classification-based image fusion methodis proposed in this paper. First, the sparse coefficients are calculated based on sparserepresentation, then using random matrix to measure it. Second, evaluate the data similarity ofthe two calculated measurement vectors. Third, adaptively select the fusion rules based on thedata similarity and fuse the two calculated measurement vectors to generate a comprehensivemeasurement vector. Finally, the fused image is reconstructed by using the measurementvector. Experiments show that the proposed method have a better performance.However, the monitoring images in WMSN are rich in edge, texture and other detailinformation, only using single type sparse basis to sparse represent may lose importantinformation. Thus, a multi-component image fusion method based on morphologicalcomponent analysis (MCA)in WMSN is proposed. First, extract the Salient Region and sparserepresent it with Curvelet basis. Second, decompose the visible image into smooth and texturecomponents with DWT and DCT. Finally, reconstruct the morphological components ofimage and added together to obtain the fused image. Experiments show that the proposedmethod have a better performance. Moreover, the proposed method has lessened the imagedata transmission, improved the effectiveness and reliability of video monitoring in WMSN.