Fusion Of Infrared And Visible Images Based On Compressed Sensing

Image fusion technology is a hot research field in recent years, the image, and the image fusion of infrared and visible light is widely used in military, remote sensing, security and video surveillance and other fields. Infrared sensor of can very good recognition and visible light sensor can provide clear scene information, through the fusion of infrared and visible light, can also use information of target heat and scene, more accurate judgment of the target, is conducive to the target identification and other subsequent processing.The traditional image fusion because all the information needs of the source image, easy storage and transmission, low efficiency of fusion. Compressed sensing(compressive sensing, CS) with relatively low sampling rate, reconstruction of the original signal, become a hot research topic in the field of image fusion. Compressed sensing in sampling and compressing the data, and the integration of only on a small amount of measurement values and instead of the whole image can evidently reduce the complexity of system storage pressure and calculation. The compressed sensing fusion, this paper mainly do the following work:(1)Gives the compressed sensing theory frame and structure, the existing fusion rules based on compressed sensing is studied, and summarizes the commonly used image fusion quality assessment index.(2)The study of CS are concentrated in DCT, based on wavelet transform, NSCT(NSCT) sparse transform. In this paper, the non sampling shearlet transform(NSST) used in compressed sensing domain, and puts forward the new compressed sensing fusion scheme, only to calculate large amount of high frequency sub with coefficient based on compressed sensing image fusion method to fuse, namely the high-frequency sub with star measurement, the measurement value is associated with the information structure; the measurement of high frequency sub band developed a new fusion rule based on spatial frequency weighted. The local standard deviation and local energy fusion guidance combined with the low frequency subband coefficients. Finally, the fusion image is obtained by inverse NSST transform. The experimental results show that with only one layer of NSST transform can reconstruct a high quality image, the fusion effect is better than other kinds of traditional compressed sensing fusion algorithm.(3)Sampling in shearlet transform(NSST) based on non compressed sensing based fusion algorithm, the fusion algorithm of compressed sensing is studied based on image segmentation. Considering the image block sampling, the lack of overall characteristics ofimage blocks between the shortcomings, resulting in image after reconstruction will produce block effect. The introduction of smooth projection in the block compressed sensing fusion algorithm(SPL) algorithm for Landweber reconstruction, effectively remove the block effect, and improve the convergence speed.