Fixdemo

[dyang37]

This PR contains the following things:

• fixes to mace3D demos and partial fix to 3D shepp Logan demo.

• Modified 3D shepp Logan generation to contain block-averaging from a higher resolution phantom (default block size 4x4x4).

  The interface is shown below:

  def gen_shepp_logan_3d(num_rows, num_cols, num_slices, block_size=(4,4,4)):
    """
    Generate a smoothed 3D Shepp Logan phantom with block-averaging strategy.
  
    The phantom will be generated by downsampling a higher resolution phantom with size (num_slices*block_size[0], num_rows*block_size[1], num_cols*block_size[2]) with block-averaging technique.
  
    Args:
        num_rows: int, number of rows.
        num_cols: int, number of cols.
        num_slices: int, number of slices.
        block_size: (int, int, int): block averaging size. A larger block_size will generate a smoother Shepp Logan phantom.
  
    Return:
        out_image: 3D array, num_slices*num_rows*num_cols
    """

• Added phantom generation functions into readthedocs.
• Re-organized 3D shepp Logan demo script to improve readability.

Tested everything with a new conda env.

[cabouman]

I ran the Sheep-Logan demo, and it works well (except for the aliasing that we are still trying to better understand.)