Research On Image Fusion Based On Multi-scale Analysis

Image fusion is an important branch of image processing.In recent years,based on multi-scale and multi-direction decomposition transform domain pixel image fusion method has become a hotspot in this field,because it can overcome the distortion of spatial domain effectively.Image fusion technology is widely used in military,civil and other fields.this paper focuses on the improvement of the shortcoming that the existing visible light and infrared image fusion algorithms cannot effectively detect infrared small target;At the same time,the correlation between the fusion coefficients of the multi-focus image fusion algorithm is improved.In the end,different fusion evaluation criteria are summarized,simulation experiments are carried out according to the standard,and the proposed improvement fusion method is verified,evaluated and analyzed.The main work and innovation points of the paper are as follows:1.Analyze the research background and significance of the field of image fusion at home and abroad,and introduce the principle and decomposition effect of the multi-scale decomposition tools,nonsubsampled Contourlet transform(NSCT),which used in this paper.Firstly analyze the research background of image fusion,research significance,present situation and development at home and abroad.Second,divide the different level according to the existing fusion algorithm,compare the advantages and disadvantages of each algorithm.Finally,the advantages of NSCT decomposition are analyzed by comparing with the common multi-scale algorithms,just like the wavelet transform and the Contourlet transform.2.In the fusion of infrared and visible images,the detection of infrared targets can be effectively detected by using wavelet packet decomposition and high order cumulant.First processing infrared image with the wavelet packet and gaussian discriminant,getting the infrared target.This method can obtain the weak target in the complex background,the target detection rate is higher,the detection result is more accurate.And then decomposing the infrared and visible images with nonsubsampled contourlet transform,getting the high frequency subband and low frequency subband.And fusion with the infrared target and background in the low frequency subband respectively,at the same time fusion the high frequency subband coefficients using regional energy and variance comparison.Finally,the high frequency subband and low frequency subband fusion coefficient are inverse transformation,getting the fusion result image.3.In order to improve the correlation of the fusion coefficient in the multi-focus image fusion technology,enhance regional information abundance,this paper proposes a method based on multi-scale decomposition and entropy rate segmentation on multi-focus image fusion.After multi-scale decomposition,edge and detail information stored in the high frequency subband,with modeling value and consistency check,can better keep more details of the image,at the same time,the low frequency subband combined with entropy rate of segmentation,assigned the coefficient to an area with the image information in closer,then fusion image according to the regional spatial frequency and regional entropy,improve the correlation between the coefficient,make the fusion image edge transition more natural.Finally,the inverse transformation was carried out on the images to get the fusion result.4.The two fusion algorithms are simulated and compared with the other four fusion algorithms.Through the comparison of subjective observation and objective fusion image standards,the validity of the proposed fusion method has been proved.The algorithm proposed in this paper can get a better fusion of visual effects and edge information,which is great significance to the subsequent image processing.