Research On CT Image Denoising Based On Conditional Generation Counter Measure Network

Compared with conventional X-ray images,CT images are clearer and have higher density resolution.Therefore,CT can more clearly display organs composed of soft tissues and pathological tissues caused by density changes in organs.However,due to the influence of CT hardware system,quantum noise and reconstruction algorithm,the reconstructed CT image inevitably has noise,which leads to the degradation of image quality.In this paper,on the basis of deep learning,CT image denoising is studied based on conditional generated countermeasure network(CGAN).(1)Build the condition generation confrontation network and study the CT image denoising method.Taking the traditional generation countermeasure network as the framework,adding specific additive Gaussian white noise(AWGN),the condition generation countermeasure network is built,and the network model is trained in tensorflow environment.The nonlinear mapping relationship between noise image and noiseless image is obtained by learning,and the influence of different loss functions on network accuracy is studied.Experimental results show that,compared with other traditional image denoising algorithms,this method can remove noise and retain feature information more effectively.(2)An improved CGAN network structure is proposed.On the basis of cgan,deconvolution network and residual network are added.Through the use of deconvolution to maintain the consistency of CT image input and output feature size,at the same time,multiple use of jump connection,effectively prevent the over fitting problem in the iterative process,more effectively retain the details in the convolution process,reduce the loss of feature information.The experimental results verify the effectiveness of the improved network.