Research Of Infrared And Visible Light Image Fusion Methods Based On Multiscale Geometric Analysis

Being a branch of multi-sensor information fusion,infrared and visible image fusion could effectively composite and extract the features of the two kinds of images.It enhances scene understanding,highlights objects,and detects targets quickly and accurately in the case of concealment,disguise or confusion.Multiscale geometric analysis(MGA)methods have the features of multiresolution,multidirection and anisotropism,which could be used to extract and analyze singularity of infrared image and visible light image,and reveal the characteristics of them.Based on infrared image and visible light image,this paper researches effective image fusion algorithms by the MGA methods,and intends to provide new approaches for infrared and visible image fusion.The paper focuses on discussing three kinds of infrared and visible light image fusion algorithm based on MGA methods.They are the noised infrared and visible image fusion method using the improved Bilateral filtering method in the Nonsubsampled Contourlet transformation(NSCT)domain,the color infrared and visible light image fusion method based on the Tetrolet transformation and the infrared and visible light image fusion method based on the Ripplet transformation.On these bases,the infrared and visible light image fusion system is explored.The researches devote to improve the visual quality of images,enhancing the scene understanding and visualization.The main research contents are as follows:(1)Researches on the noised infrared and visible light image fusion method based on the NSCT and improved Bilateral filtering method.In order to solve the edges loss problem while reducing the noises in the noised infrared and visible light image fusion algorithm,the NSCT and improved Bilateral filtering method was combine.The infrared image and visible light image were decomposed respectively by NSCT firstly.Then carry the edge detection on the output high frequency information by improved Bilteral filtering method,with the help of its features that it can protect edges while filtering.Next,the detailed image of the visible image high frequency components was added to the visible image high freqence coefficients,promoting the edge extraction capacity of the algorithm.In low frequency region,this algorithm adopted the standard variance coefficient of partial region as the fusion rules.Finally,it got the reconstructed image by doing the inverse NSCT transform on the fused high frequency and low frequency subband coefficients.The experiment results show that compared with traditional fusion methods,this algorithm can effectively reduce the noises in the images and enhance the linear features extraction ability.(2)Researches on the color infrared and visible light images fusion by the Tetrolet transformation.In order to solve the color distortion problem in color images fusion algorithm,the Tetrolet transformation and l?? color space method was adopted.The Tetrolet transformation decomposition framework was improved according to the infrared image and visible light image features firstly.In the fusion process,the visible light image was converted to the l?? color space to get three irrelevant color channels in order to process the gray channel image and color channel images respectively.Secondly,the gray channel image l and infrared image were decomposed by the Tetrolet transform.The neighborhood energy and proximity were introduced to the low-pass coefficient fusion rule.The Tetrolet coefficient was fused by the compressed sensing method.The fused gray image was got after the Tetrolet reconstruction.Finally,the final fused image was obtained by mapping the gray image and ?,? color images to the RGB color space.The experiment results show that this method could get fusion color image with abundant linear features and solve the high color distortion problem.(3)Researches on the image fusion method by Ripplet transformation.In order to solve the problem that the edge details in the original images could not be completely extracted to the fused image,this paper adopted a novel Ripplet transformation image fusion algorithm.Firstly,this algorithm got the decomposed high frequency and low frequency subband coefficients by the discrete Ripplet transform.For the low frequency subband coefficients,it adopted the 3-layer NSCT to do coefficients fusion.For the high frequency subband coefficients,calculate the weight coefficients of phase equalization,articulation and gray scale information,and get the weights of the high frequency coefficients by the activeness of the coefficients.The fusion high frequency coefficients are acquired according to the weights.Finally,the inverse Ripplet transform was used to get the final fused images.The experiment results show that this algorithm could better keep the images edge contour and improve the contrast of the clear part.4)Design of the infrared and visible light image fusion system.Based on the earlier theoretical study,a new kind of infrared and visible light image fusion system was designed.On the software platform,verify the feasibility and efficiency of the video preprocessing,video conventional processing and video fusion algorithm,and confirm optimal algorithm and its processes.On the hardware platform,the cameras and capture card were used to establish real time image processing system on the Personal Computer(PC)platform.Then compile the optimized algorithm to realize the algorithm transplant and firmware.Based on the MGA theory,the paper proposed several kinds of infrared and visible light image fusion algorithm.It abundants the infrared and visible light image fusion system.On this basis,a kind of infrared and visible image fusion system was established,supplying foundamention for practical application.