Research On Image Denoising And Separation Based On K-edge Image

K-edge imaging is a promising technology extension of multispectral CT,which is characterized by the drastic changes in the attenuation coefficient over a range of energies.Based on this feature,K-edge imaging technology obtains clear,high-quality images by shooting the target material before and after K-edge energy.K-edge imaging technology can effectively differentiate the different soft tissues of human body,remove beam hardening artifacts and reduce radiation dose,which is a promising medical imaging method.In this paper,Independent Component Analysis(ICA)is used to remove the noise from the K-edge images.In the simulation experiment of noise reduction for K-edge image,the input sparse signal is denoised by setting global threshold function to preserve the useful information.With a fast ICA method,ie FastICA,the denoised image has a small amount of noise and a good visual effect,which is well adapted to the K-edge characteristic image.In fact,both the region of interest(ROI)and the model of the K-edge feature are indistinguishable.Since the number of images is large and the model is simple,the K-edge image can be segmented by using a Support Vector Machine(SVM),and SVM with the optimized parameters.In SVM algorithm,the target image is sampled to select the training set.Then the SVM-based model is used to train the samples.Finally,the model is used to predict and determine the target image,and finally the image segmentation is achieved.Through the experimental analysis,the optimal separation algorithm is obtained.The accuracy of the segmentation of the target image is expected to reach 100%;including the artificial sampling,the whole algorithm can be completed within 8 seconds,which can meet research conditions of the images of K-edge characteristic composite models.