Research On Image Fusion Algorithm Based On Compressed Sensing

Image fusion is one of the most important part of image processing.It has been widely used in remotely sensing image,computer vision and medical image processing region.However,because of the increase of realistic demand,image resolution and signal sampling frequency become huge at the same time which increases the cost of signal processing,transmission and storage.Compressed sensing theory breaks through the frequency limitation of traditional Nyquist sampling theory.Under this compressed sensing structure,the cost of signal sampling and reconstruction decreases heavily.Meantime,the compressed sensing theory has broadened the horizon of image fusion region.Compressed sensing theory mainly contains three parts.They are the design of sparse transformation matrix,the construction of measurement matrix,and the design of reconstruction algorithm.It is of vital importance to design a quick and accurate reconstruction algorithm and apply it to image fusion region.The mainly researching and innovative work of this dissertation is concentrating on those aspects.(1)The first part of this dissertation is describing the development status at home and abroad of compressed sensing and image fusion.Then it describes the three basic steps of compressed sensing theory.Meantime,it has analyzed the important parameters of different methods and it gives the qualitative analysis from objective and subjective perspective.(2)It introduces the basic theory of current image fusion method and it describes the fusion method on spatial domain.Then,it emphasizes the main method of image decomposition in transform domain.Finally,it tells the physical significance and computing method of those evaluation index.(3)In order to design a qualified algorithm which owns an excellent fusion effect while keeping a high reconstruction probability.Based on generalized orthogonal matching pursuit algorithm,an optimized generalized orthogonal matching pursuit algorithm comes up.Comparing with orthogonal matching pursuit algorithm,stagewise orthogonal matching pursuit algorithm and compressed sensing matching pursuit algorithm,this method has adopted recall feature and it changes the way of chosing atoms.So it becomes running faster,reconstruction precision is higher and more flexible.Besides,it has considered the relationship between different atoms which can reduce the interference of vertical direction.Furthermore,infrared image enhancement module is added into this structure which improves the performance of total algorithm.(4)In order to match the subjective and objective assessment criteria.After the description of the principle and method of non-subsampled contourlet transform.The high frequency fusion method is optimized,and this algorithm becomes suitable for human's eyes.The structural similarity coefficient is used here.Meantime,according to the simulation experiment,the feasibility and stability of the algorithm are verified from subjective and objective aspects.