Research On Multiplicative Noise Image Processing

Image processing plays an important role in the areas of our life. As is well known, multiplicative noises are commonly found in many real world image processing applications, such as SAR images, microscope images and ultrasonic images. Unlike additive noises, these noises are much more difficult to be removed from the corrupted image, mainly not only because of their multiplicative nature, but also because of their distributions which are generally not Gaussian.In this thesis, we are mainly concerned with multiplicative denoising problem and segmentation method. The image denoising method:we consider a hybrid method for removing multiplicative noise e.g. speckle noise. Our model consists of l1data-fidelity term and the nonlocal total variation as regularizer. The h data-fidelity term can preserve edges during despecking framework in the curvelet domain. We import the nonlocal total variation as regularizer which can recover the textures and local geometry structures. Moreover, the efficiency of the algorithm adopted here is based on operator Augmented Lagrangian for the hybrid method. Experiments show that the proposed scheme outperforms the most recent methods in this field.The image segmentation part:the image segmentation model based on the level set method is introduced. Image segmentation is a fundamental problem for image processing. The problem of image contains multiplicative noise, which makes it hard in the field of image segmentation. We draw our inspiration from the modeling of multiplicative noise. By using a MAP estimator, we can derive a variational level set approach whose miniminzation corresponds to the multiplicative noise image segmentation. The new variation method with a non-convex regularization term has advantages over classical regularization term in level set methods. Experimental results based on both synthetic and real SAR images show that this method has a better segmentation effect.