Super-resolution PET image reconstruction with sparse representation

In the positron emission tomography (PET) field, reconstructed images are often blurry and contain noise. These problems are mainly caused by the low-count problem. As sparse technology becomes more widely used, sparse prediction is increasingly prone to be selected to solve the problem. In this paper, we propose a new sparse prior method to process Low-resolution PET reconstructed images. In the proposed strategy, two dictionaries (D
1
for low-resolution PET images and D2 for high-resolution PET images) are trained from numerous real PET image patches. Subsequently, D
1
is used to obtain the sparse representation for each patch of the input PET image. Finally, a high-resolution PET image is generated from this sparse representation using D2. The results of experiments indicate that the proposed method exhibits stable and superior effects that enhance image resolution, detail recovery. Quantitatively, this method achieves a better performance than traditional methods in terms of root mean square error (RMSE). The proposed strategy provides a new and efficient approach to improve the image quality of reconstructed PET images.