FBP: process stack of sinograms
Process stack of sinograms instead of a single sinogram. Two reasons:
- Be consistent with the other processing steps
- Performance (see [1])
The Cuda kernel will have to be modified accordingly. Perhaps it would be best to create one kernel for "single sinogram", and another for "sinogram stack".
The Cuda kernel would not necessarily process all the stack at once, but for example a fixed number of sinograms (ex. 4).
[1] https://link.springer.com/article/10.1007/s11554-019-00883-w