Читать книгу Machine Learning for Tomographic Imaging - Professor Ge Wang - Страница 31

It is suggested that dictionary learning for a sparse representation effectively mimics the HVS perceiving structural/semantic information in natural images. In this section, we will use low-dose CT reconstruction as an example to illustrate how the dictionary learning method is applied in the CT field. This case is described in two parts. In the first part, we will introduce a statistic iterative reconstruction framework (SIR) extensively used in the CT field, which exemplifies the application of the Bayesian approach for CT imaging. In the second part, based on SIR we will illustrate the application of dictionary learning for CT reconstruction to preserve subtle features in reconstructed images from low x-ray dose data. This example will show the power of dictionary learning as a machine learning technique.