The Study Of Multi Focus Image Fusion Based On Multi-scale

Image fusion has a wide range of applications in military and medical fields.Image fusion refers to the fusion of two or more images into a new image by using related methods.Multi-focus image fusion technology is an important branch in the field of image fusion.It is used to solve the problem that the optical sensor can not ensure that all the objects in the scene are clear due to the difference of the depth of field when the scene is shot by a certain sensor.Multiscale is a common method of multi-focus image fusion,among which the wavelet transform and Contourlet transform are the most representative.As a multi-scale decomposition method of image,wavelet transform has better spatial and local characteristics and is widely used in image processing.The down sampling in the discrete wavelet transform image fusion algorithm discards half of the filtered coefficients,resulting in the lack of translational invariance,which makes the fusion algorithm vulnerable to mis-registration of source images.Therefore,the image fusion algorithm based on non-subsampled discrete wavelet transform has been applied.The corresponding fusion rules are adopted,and the consistency of the fusion coefficients is detected to make up for the lack of subsampled and improve the fusion effect.In this paper,based on the discrete wavelet transform and non-subsampled discrete wavelet algorithm based on the image multi-focus fusion simulation study shows that the non-subsampled discrete wavelet transform algorithm with translational invariance,but also has high fusion image mutual information,the fusion effect is better than the discrete wavelet transform image fusion algorithm.An adaptive algorithm is proposed for the fusion rules.By comparing the fusion experiment can prove the effectiveness of the adaptive algorithm on the fusion.The wavelet transform is suitable for representing the same-sex singularity features,but not for the singularity features with anisotropy in images.Based on this problem,the Contourlet transform is introduced.Contourlet transform notonly contains good local characteristics of time-frequency,but also contains multi-directional and anisotropic and can achieve multi-scale,multi-directional decomposition of the image.However,Contourlet transform lacks translational invariance,so non-subsampled Contourlet transform is applied to image fusion.The non-subsampled Contourlet transform not only includes the advantages of Contourlet transform but also has the invariance of translation.The experimental results show that the multifocus image fusion based on non-subsampled Contourlet transform is better than the non-subsampled discrete wavelet transform.In addition,an improved adaptive fusion rule is proposed for the fusion rule.The effectiveness of the improved adaptive fusion rule can be verified through comparative experiments.