Spike Noise Removal For Magnetic Resonance Imaging Based On The Information In The Background Region

Magnetic Resonance Imaging has become one of the most important clinical imaging method. In magnetic resonance imaging, the raw data will be damaged by some external influences, such as hardware failure and patient movement, resulted to reconstructed images with artifacts and affect the image quality seriously. Many algorithms have been proposed to deal with artifacts in MR images, the results were well. In this thesis, the theory of removing spike noise is introduced and several existing removing spike noise algorithms are described and analyzed. We proposed a new approach to locate and remove spike noises based on the information in the background region of the corrupted image. Our algorithm can eliminate spike noise of various modes, and can be better able to handle a large number of consecutive spike noises in the K-space, eliminating image artifacts effectively compared with existing algorithms. Using image processing algorithms to remove spike noise in the K-space, almost no damage to the image quality without additional scanning time, both to improve the patient feeling, and also can improve the utilization of the equipment effectively.The3D magnetic resonance imaging is more popular now, so the algorithm is extended to handle3D data, so there are more spike noise artifacts of different modes can be removed in magnetic resonance imaging.There are different shapes background for different parts of the MR images, in order to extract more background information, the image background is selected interactively. So we can select any shape of the background region in different modes of magnetic resonance images to remove the spike noise.