Research And Realization Of CT Image Super-resolution Restoration

In the process of recording a digital image, there is a nature loss of spatial resolution caused by the influence of motion blur, point spread function and system noise. Super resolution reconstruction is to reconstruct a high resolution image from a serial of low resolution images. It can promote the quality of the degraded image. This technology is widely used for remote-sensing, medical imaging and high-definition television standard and is of great significance for academic study.This paper aims at rising the resolution of CT image, the main focus is on POCS(projection onto convex set) algorithm which is based on set theory and MAP(maximum a posterior) algorithm which is based on Statistics and Probability Theory. The related research such as imaging model, regularization of the ill problem and MTF theory is also introduced.For the classical POCS algorithm, although it has strong ability to contain different prior information, it has some difficulty in fusing the detail information of different images in the iteration process. Basing on the classical POCS algorithm, we put forward a method that combines POCS algorithm and wavelet fusion to reconstruct a high resolution image. For MAP algorithm, although it has strong super resolution restoration ability , it has some difficulty in suppressing the oscillatory artifact and the noise. In the condition of low SNR(singnal to noise ratio), the restoration is influenced badly. Markov random field is introduced based on MAP frame and the limited energy in neighborhood is added to the process of MAP iteration to constraint the oscillatory artifact and noise.In the experiment, the POCS and MAP algorithm is performed directly on CT image. The algorithm is estimated by comparing the difference of the MTF. Then in the simulation experiment, the degraded image is acquired by blurring and adding noise to estimate the restoration performance under different condition.Experiments demonstrate that wavelet based POCS algorithm can improve the ability of fusing different information, and the detail of the image is more prominent, the image quality is better. For MAP algorithm, the Markov random field can constraint the oscillatory artifact and the noise the image quality is improved.