Multi-spectral Image Fusion Method Based On Non-separable Wavelets And Directional Filter

In the last century,the concept of multi-sensor was been proposed and it had developed for a long time.Image fusion technology had been greatly developed with the rapid development of computer information technology.Its applications are becoming more and more widespread.After entering into the new century,image fusion is becoming a hot research topic.Image fusion has a wide range of applications in non-military areas and military fields,for automatic target recognition,machine vision,remote sensing,robot,complex intelligent manufacturing systems,medical image.Image fusion is generally divided into three levels,pixel-level fusion,feature-level fusion,decision-level fusion.Image fusion is a technique,which multiple imaging of the same scene,region and the same target by single sensor or imaging of the same scene,region and the same target by multi-sensor,then dealing with these imaging through certain rules,lastly obtain a new image.Such images or image features obtained are more accurate and more completed than a single image in terms of various indices and information.Because a image obtained by a single sensor has some differences and limitations in all aspects of information.Through image fusion processing,the information of some single sensors may be integrated,then we can make information data errors minimized and get some image information what more accurate and more completed and perfect.We can get more comprehensive,more accurate and reliable information of the same scene,region and the same target by merging the information of many sensors.And the image dealed with image fusion technology is more adapted to the visual characteristics of human beings and machines,then it becomes relatively simple and easy to operate which further in-depth treatment of image results.The wavelet transform that applied to image fusion usually refers to the two-dimensional wavelet transform and the one-dimensional wavelet transform is usually used to one-dimensional signal transform.In the field of image fusion,the wavelet transform has been widely used.The two-dimensional wavelet of the two-dimensional wavelet transform generally refers to the traditional two-dimensional tensor product wavelet,and the wavelet space of this wavelet can be decomposed into the tensor product of two one-dimensonal wavelet space.The traditional tensor product wavelet has some defects of the directionality,then it can only capture the horizontal direction information,the vertical direction information and the diagonal direction information.And the traditional tensor product wavelet can not take full advantage of the unique characteristics of the data itself and it is not the best representation of function The defects of the directionality of the traditional tensor product wavelet lets itself not make full use of the geometric regularity of the image.Compared with the traditional tensor product wavelet,the development of the non-separable wavelet is far behind but it has become a hot topic for researchers.Compared to the tensor product wavelet,the non-separable wavelet can capture all directions and it has all the ability to multi-resolution of the traditional tensor product wavelet.For the low computational complexity of the traditional tensor product wavelet,the non-separable wavelet has higher computational complexity.The wavelet transform with the non-separable wavelet can obtain more high-frequency information,then it is more conducive to the analysis of image details.This paper puts forward a multi-spectral image fusion method based on non-separable wavelets and directional filter.According to the theory and the method of two-dimensional non-separable wavelet filter banks,this paper focuses on the two-dimensional structure of a four-channel non-separable wavelet filter banks.Then use it doing the experiment of image fusion with the directional filter.In the fusion process,this paper has a average treatment for the coefficient of low frequency and a treatment which has larger absolute value for the coefficient of high frequency.And use the traditional IHS method,the IHS-WBCT method and the.IHS-DWT method doing the image fusion experiments with the same experimental images in the same experimental environment.Evaluate and analysis all the results and compare the results.The experimental results show that compared with the other three methods,the method of this paper can maintain higher spatial resolution and maintain better overall spectral information of the image.