Abstract:

Density reconstruction of non-axisymmetric object from few projected data needs to be studied in high-energy flash radiography. The existing kinds of TV algorithms which exploits the idea of compressed sensing consider the local similarity of images, but they don’t consider the non-local similarity. In view of these problems, this study proposes a total variational reconstruction technique TV-GSR based on group sparse regularization. This technology integrates the group sparse model into the TV framework, and considers the local similarity and non-local self-similarity of the object image, making full use of the prior sparse information of the image. Besides, the proposed technology also uses the four-point symmetry of up, down, left and right of object to reduce the size of image reconstruction. Thus, the reconstruction accuracy increases and the reconstruction speed accelerates. Simulation experiments show that the proposed TV-GSR algorithm improves the reconstruction accuracy of images in noiseless and noisy scenarios, and has good effects on high-energy flash images and CT images with rich texture details, and is universal.

Key words: flash radiography, density reconstruction, image reconstruction, compressed sensing, group-sparsity regularization, total variation, sparse information, similarity