Learning Tomography: How many projections?

Two parameters are crucial to compute a good reconstruction: the  number of projections and the angular coverage. Through several examples, we'll see their impact on the reconstructed image.


1- How many projections do we need?

In the previous posts, 180 1D-projections were used for computing a reconstruction corresponding to data collected with a step of 1° over an angular range of [0-180[. Is it enough? Do we need more or less projections?

Modify the number of projections used to compute a sinogram by changing the line 'var nbProj=180;' of the script sinogram.js.

1. Compute different sinograms from Lena (or what you want) by choosing 36, 90, 180, and 360 projections (corresponding respectively to an angular increment of 5°, 2°, 1°, and 0.5°). 
2. For each sinogram, compute the weighted back-projection reconstruction (ramp filtration and reconstruction). 

You'll get something comparable to Fig. 1.

Fig.1: 2D reconstructions calculated from sinograms of 5°, 2°, 1°, and 0.5° angle step, respectively. Bottom: Zoomed area above Lena's hat. The moire pattern fades gradually as the projections number increases.

W O R K    I N   P R O G R E S S

A moire pattern is clearly visible when the number of projections is decreasing.



>>> In a next post: The impact of the angular coverage [Link]

Other crazybiocomputing posts

Further readings are available in ...

• Learning Tomography Series  [Link]
• Image Processing TOC [Link]