Multi-focus Image Fusion Method Research Based On Two-channel Non-separable Wavelet

Image fusion means that in accordance with certain rules,making the same scene or target imaging of multiple sensors or multiple imaging of a single sensor process,and then generate a new image,and the result of the new fusion image must be better than the original image,that is,from subjective visual effect and objective index evaluation standards are all optimal.The data and information of the fusion images that are selected are usually insufficient,and the target scenario information is not comprehensive,and so on.The fused images have overcome these defects in a certain extent,which are more conducive to the deep analysis of the image application and more occasions.Along with the development of the times,image fusion has been widely used in the field of digital images,remote sensing images,medical images and other non-military fields and military fields,which is becoming a researching hotspot.According to the information abstraction level of image fusion,it can be divided into three levels from low to high,pixel level,feature level and decision level.For the methods of multi-focus image fusion,it has existed some mature theoretical support.At present,wavelet transform has been very widely used in image fusion filed.but traditional tensor product wavelet transform can not make full use of the data itself unique characteristics,and it is not the optimal representation method,besides,it can only capture limited directional information:horizontal,vertical and diagonal direction.The lack of directions makes it can’t make full use of the geometric regularity of image itself.Multi-focus image fusion means making multiple target image from the same scene twice or more times of imaging process by using computer aided by related algorithm,to make the clear of the multiframe imaging area form a new image.In terms of research in the field of image processing of the fused images,the fused image has more complete data information what we need,which helps us with in-depth analysis of the image data,and accurate recognition and more thorough understanding of the imaging target.In the field of image fusion,the common methods of multi-focus image fusion are based on wavelet transform fusion method,based on block fusion method,based on pyramid decomposition fusion method and so on.These several methods will be introduced in this article,and also making corresponding analysis and evaluation of the fusion results.With the development of science and technology,science and technology upgrading of equipment and the related applications in the field of image processing,some new methods of multi-focus image fusion are proposed.They are based on lifting wavelet transform multi-focus image fusion method,based on Curvelet fusion method,based on morphological wavelet fusion method and so on.Because these new methods are not very mature application, they will not be introduced in this article.In the field of image processing,tensor product wavelet only emphasizes the vertical and horizontal direction information.Aiming at this defect,this paper proposes a new multi-focus image fusion method,based on the two channel non-separable wavelet multi-focus image fusion method.The method researches the structure of the non-separable wavelet filter bank,producing a series of2×2filter bank,then making the multi-focus image sample under multi-scale decomposition with the constructed non-separable wavelet filter bank. Image fusion was performed after decomposition of the original image using low frequency component of the coefficient values in small and high frequency component coefficient absolute value in large fusion rules. From a series of relevant experiments based on the fusion rules,by evaluating and analysising the results of the experiments,prove that the method has better fusion effect than other fusion methods.The result of the fused image has rich edge information,higher resolution and spatial resolution,and its convergence performance is better than the fusion performance based on the tensor product of discrete wavelet transform (DWT) fusion method,proving the feasibility and accuracy of this method.