Superresolution Reconstruction Methods

It is generally that the image resolution is determined by the system. However, in practical application, in order to get more detailed images, people try to reduce the X-ray focal spot size or the detector size to enhance the resolution of images. However, if the X-ray focal spot is small, on the one hand, the penetrability of X-ray is degraded, on the other hand, the detector becomes inefficient, consequently we must prolong the scan time to ensure the detector efficiency. If the detector cell is small, not only the cost is expensive, but also the detector is inefficient, in order to ensure the detector efficiency, we must prolong the scan time or increase X-ray focal spot size. In a way we can obtain a little better results by the two methods, but it is too limited. Well than, under the existing hardware conditions, how to break through the inherent resolution and get high-resolution images (that is, super-resolution imaging) becomes one of the most noticeable problems.In this paper, we study super-resolution imaging problems. Based on the characteristics of CT platform in general, we propose two kinds of super-resolution methods, i.e. platform offset method and platform offset, rotation method. Concretely, demodulate a set of high resolution projection data from two sets of low resolution projection data by SIRT and moderate smoothness method, then reconstruct a high resolution CT image by the fan beam FBP algorithm. The simulation results verify the validity of the two scanning modes and reconstruction algorithms, and the resolution of CT image corresponding to the second scanning mode is obviously better than that corresponding to the first one. Besides, we also give the reconstructed CT images with offset and rotation errors. At last, we consider the geometry of a scanner. The geometry includes the size of the detector cells, the size of the X-ray focal spot, the distance between the detector and the source, and the distance between the source and the iso-center. It is theoretically that the geometric parameters' influence on the reconstructed images is explained, consequently we ravel the conditions of improving super-resolution imaging resolution by overlap sampling.