Abstract:

When Wireless Multimedia Sensor Network (WMSN) is used for environment detection and the infrared image and the visible image collected from the same scene are fused, the traditional approaches have large amount of fused data and do not fully consider the internal sparsity and the complexity of structure features, so the fusion quality is low.The theory of sparse representation is applied to WMSN infrared and visible image fusion. Based on the original DCT redundant dictionaries, the KSVD training method is combined with the Simultaneous Orthogonal Matching Pursuit (SOMP) algorithm to do effective sparse representation for WMSN infrared images and visible images.And adaptive weighted average fusion rule is selected to deal with the sparse representation coefficients.Experimental results show that,compared with traditional infrared and visible image fusion methods based on spatial and transformed domains, the proposed method can effectively preserve the useful information and get the better fused image.
      

Key words: WMSN;image fusion;sparse representation;simultaneous orthogonal matching pursuit;adaptive weighted average