In order to reduce the noise of coronary angiography image sequence, enhance the diagnostic accuracy for coronary heart disease, and eventually acquire superior image quality under low X-ray dose, a method of spatial-temporal filtering for coronary angiography images was proposed. By introducing the idea of threshold noising in wavelet denoising into the Fast Discrete Orthonormal Stockwell Transform (FDOST), a soft-threshold denoising algorithm based on FDOST was proposed for the spatial denoising of coronary angiography images. The conventional wavelet denoising was used for temporal denoising of coronary angiography images, taking advantage of its time smoothing feature. Hessian matrix was used in pre-processing to track the line-like structure of coronary angiography images. The simulation and experimental results show that the signal-to-noise ratio and contrast-to-noise ratio of the denoised images are improved significantly compared with the original image, and the proposed method is suitable for the denoising of low-dose coronary angiography image sequence.

The X-ray phase contrast Computed Tomography (CT) can produce high contrast images by the X-ray phase information alteration, which comes forth after the X-ray passes through the sample, and it is highly favorable to the imaging of light elements and can get much higher contrast resolution than the absorption contrast CT. Grating-based Differential Phase Contrast CT (DPC-CT) shows great clinical prospects due to the possibility of using a conventional X-ray source, but the X-ray radiation dose issue limits its clinical applications. Concerning such inadequacies, an image reconstruction method for DPC-CT named DD-L1 was proposed. This algorithm combined Compressive Sensing (CS) theory with CT iterative reconstruction technique and introduced distance driven forward and backward projection computation strategy. The experimental results show that DD-L1 algorithm can generate tomographic images of higher quality even when the projection data is incomplete.