| In recent years,as the information fusion technology research results on the basis of rapid progress,image fusion technology has gradually developed.The purpose of image fusion is to obtain multiple images through multi-sensor,and get the new target image after all the source images are processed by some fusion algorithms and rules.Image fusion is not only used in medicine,remote sensing,environmental protection and other fields,but also radar images,anti-terrorism and other military applications are also widely used.At present,because Wavelet transform has good localization property in frequency domain and time domain,it has become more research methods in recent years.However,Wavelet can‘t express the characteristics of image anisotropy because it can only express zero dimensional singularity and can‘t express the limitation of higher dimensional feature.After recognizing the shortcomings of Wavelet transform in multi-focus image fusion,a new multiple scale transform Curvelet transform,which is developed by Wavelet transform and Ridgelet transform and which compensates Wavelet for missing edge description,is proposed.Since the Curvelet transform can sparsely represent the smoothing and edge parts of the source image,the twodimensional signal description of the straight lines and planes can be well performed.And the Curvelet transform can describe the details of edge and curve in the finished image with excellent directionality and anisotropy.The different criterions are applied to high and low frequency parts processing respectively.Because of these superior properties,Curvelet can be widely used in image fusion,desorption enhancement and feature extraction,and the effect is remarkable.In this paper,the traditional low-frequency and high-frequency coefficient fusion rules are improved to make them more suitable for Curvelet transform.Compared with other algorithms,the results show that the imageprocessing algorithm in this paper has more excellent results. |
PDF Full Text Request