Image Fusion Research Based On Compressive Sensing

With the continuous development of information science and technology,people’s requirement for the effect of image fusion is higher and higher.The study on the algorithm of image fusion has been one of the highlights in the image processing. And the quality and effects of image fusion will do great benefit for more in-depth research,understanding and application. At present,the image fusion has a wide range of application in satellite remote sensing image processing,computer vision,video processing,medical image processing and the military field.Compressive Sensing Theory breaks through the restrict from Nyquist sampling theorem.The original signal can be reconstructed accurately or approximate inexactly with the sampling frequency that is much less than nyquist sampling frequency by extracting comparatively few signal measured values randomly under some conditions that the signal is compressible or sparsity.First of all,this paper expounds the background and the significance of the research. Secondly,it describes the research status of image fusion technology and Compressive Sensing Theory at home and abroad.Thirdly,it introduces the theory’s architecture of image fusion technology,including the level of the image fusion classification and their corresponding advantages and disadvantages,the process of the image fusion,the guidelines that should be followed,the image fusion algorithms which are commonly used and evaluation of fusion image that are both subjective and objective.it imports the Compressive Sensing Theory after analyze the problem comes from the process of compression of the traditional sampling based on the traditional image fusion technology and introduces the theory briefly.Then it describes,analyze and summarize the theoretical framework,including the selection of sparse matrix,signal sparse,selection of the measurement matrix and the signal reconstruction.This paper proposes an improved compressive sensing image fusion algorithm based on both the regional gradient method and the regional energy method.Firstly,a orthogonal wavelet transform matrix should be constructed,and we should use wavelet transform to make image sparse. Secondly,hadamard matrix is used as the measurement matrix to get measurement value of sparse signal,and then gradient values and energy values are calculated corresponding to the measured values.Thirdly,we fuse the measurement values according to absolute value method and weighted average method.Finally, image is constructed according to the measured value of fused.The experiments’results show that this algorithm can obtain better effect of image fusion compared to single regional gradient method and regional energy method under different sampling rate.