Multi-focus Image Fusion Based On Non-separable Additive Wavelet

Multi-focus image fusion(MFIF)is one of important research content in the field of image fusion.MFIF is defined by using a computer combined with a specific algorithm to fuse multiple images from two or more sensors to different focal points of the same scene,to combine the clear regions of these images,a better result image is obtained.It overcomes the defects caused by the focusing of the optical lens by processing the image so that objects of different distances in the same scene can be clearly displayed in one image.That lays the foundation for subsequent image processing(such as edge feature extraction,pattern recognition,etc.).At present,the most widely method of fusion for multi-focus images is a fusion method using multi-scale transform,a method based on pyramid decomposition and a method based on wavelet transform.Among them,non-separable wavelet is isotropic and can improve the resolution of the fusion result image effectively.However,when the image is decomposed by non-separable wavelet algorithm,a large number of convolution operations are introduced,which leads to a large amount of computation,and some image information will be lost when the image is reconstructed.The additive wavelet only needs to add and subtract in image reconstruction,which makes up for the non-separable wavelet defect.Therefore,the convolution factor in the traditional additive wavelet decomposition algorithm is replaced by the non-separable wavelet low-pass filter,and the non-separable additive wavelet algorithm is proposed,and the algorithm is used in the multi-focus image fusion.The research work of this paper is as follows:The two-dimensional non-separable wavelet transform is studied.By the method of constructing the high-dimensional non-separable wavelet filter bank,a two-channel non-separable wavelet filter bank,a three-channel non-separable wavelet filter bank and a four-channel non-separable wavelet filter bank are constructed.A new multi-scale and multi-resolution analysis method for images is proposed.The tradition non-separable wavelet is based on Mallat algorithm for image decomposition,a large number of convolution operations are performed,and the algorithm has a large amount of computation and some image information will be lost.Therefore,a non-separable additive wavelet is constructed to solve the defects caused by the traditional non-separable wavelet.The non-separable additive wavelet is applied to multi-focus image fusion,a multi-focus image fusion algorithm based on two-channel,three-channel and four-channel non-separable additive wavelets is proposed.Firstly,the algorithm uses the low-pass filter of constructed non-separable additive wavelet filter banks to decompose the original images,then the corresponding low-frequency sub-images and wavelet plane series images(high-frequency sub-images)are obtained.Secondly,the appropriate fusion rules are selected to fuse the wavelet plane series images and original images.Finally,image reconstruction is performed to obtain a result image.By comparing the proposed method with the method based on tenser product wavelet,the method based on Laplacian pyramid and the method based on three-channel non-separable symmetric wavelet,the results show that fusion image by proposed method has the best visual effects and the highest image definition and spatial frequencies.