A Research Into Image Denoising And Cerebral Vascular Enhancement

Image denoising has always been a hot topic in image processing area.In a variety of applications,such as astronomy,video surveillance,3D reconstruction,image segmentation,edge detection,remotely-sensed image and many others,due to environment disturbance and the inherent fault of imaging devices,images collected are always with different density of noise.Before further processing of these images,it is necessary to denoise them to improve the image quality.There are many kinds of noise such as impulse noise,Poisson noise,Gaussian noise,mixed noise and so on.To denoise different kinds of noise,a certain denoising method is needed.These algorithms usually base on a certain kind of noise model and then are designed based on some assumption.Poisson noise and Gaussian noise are two most common noise models in denoisng field.In this thesis,we did some research of both noise models.We proposed a Poisson denoisng method based on BLS-GSM method.For Gaussian noise,we proposed a new strategy based on BM3 D and accelerated the algorithm with OpenMP.MRA vascular enhancement technology plays a very important role in the precaution,diagnosis and treatment of cardiovascular and cerebrovascular diseases.In this thesis,we focused on the application of grey 3D SUSAN operator in MRA brain vascular enhancement.We firstly denoised the original data.And next we enhanced the brain vascular based on 3D SUSAN operator.Finally we displayed the enhanced brain vascular via volume rendering in 3D.The experiments show that the enhancement method is effective.In this thesis,we introduced the 3D SUSAN operator which has strong anti-noise ability into the direct volume rendering transfer function to construct a new 2D transfer function space,i.e.grey 3D SUSAN operator feature space.Because 3D SUSAN operator is not sensitive to local noise at all and has strong anti-noise ability.Thus,by enhancing brain vascular with the method proposed in this thesis,we can reach better displaying effects than traditional 2D feature space.