Hamming OMP spot shape

[paulshuker]

• We could potentially simplify the OMP scoring process by not computing a spot shape using coefficients for every dataset individually.

• Instead we could use a Hamming window.

• The Hamming will have radius in y/x and radius in z specified by hamming_yx and hamming_z. They can default to the same values as radius_yx and radius_z respectively which are already in OMP for spot detection.

• The Hamming will be

  # sz_y x sz_x x sz_z
  np.hamming(radius_yx)[:, np.newaxis, np.newaxis] * np.hamming(radius_yx)[np.newaxis, :, np.newaxis] * np.hamming(radius_z)[np.newaxis, np.newaxis, :]

• This Hamming will then become our kernel which convolves with every OMP coefficient (after the coefficient is non-linear functioned, see #107).

[paulshuker]

After looking at multiple datasets, the OMP spot shape (i.e. the sizing) does vary between datasets. Therefore, there is reason to believe that we should not move to a fixed Hamming spot shape. This issue will be closed for now.

[paulshuker]

Now that I have completed issue #262, this may need looking at again.