Research And Implementation Of Multi-source Image Fusion Algorithm Based On Multiscale Transformation

Since the human eye's ability to recognize grayscale is much lower than that of color recognition,pseudo-color processing of grayscale images during symptom diagnosis and treatment allows doctors to obtain more and more accurate information.Combining the non-subsampled contourlet transform(NSCT)with translation invariance,multi-scale,multi-directional characteristics,and pulse-synchronization and acquisition characteristics of pulse coupled neural networks(PCNN),This paper improves the medical image fusion algorithm.The image fusion algorithm based on NSCT and PCNN is mainly studied.Reasonable and effective fusion rules for low-frequency subband and high-frequency subband are proposed respectively,and pseudo-color processing of gray-scale fused images.The purpose is to make the fused image retain more energy information,contour information,detail information and source information of the source image.More convenient for human observation.First of all,the registered source image is decomposed by NSCT to obtain low-frequency subbands and several high-frequency subbands.According to the characteristics of energy information and contour information concentrated in the low frequency sub-band,the low frequency sub-band is combined by the combination of gray value and gradient value.And then,According to the characteristics of the detailed information in the high frequency sub-band,the high frequency sub-band is improved by using the improved PCNN.Finally,the source image common area is calculated based on the gray scale information of the source image.The source image is processed using NSCT and PCNN to obtain a grayscale fused image and fused image source information.According to the human visual characteristics,the pseudo-color processing is performed on the gray-scale fused image using the common image of the source image and the fused image source information to obtain a pseudo-color fused image.To verify the applicability and effectiveness of the proposed algorithm.The simulation experiment was carried out by using multiple sets of medical CT and MRI images.The fusion results show that the proposed fusion algorithm is generally superior to other image fusion algorithms,and is more in line with the human visual system,which is conducive to the discovery of lesions.