• Median filter test on the Cd spectrum image at 1.50mm with size=4:
  
• Top images are pre filter, bottom are post filter.
• Spectrums are the number of photoelectrons as a function of the pixel position
• Adjusting the size to 1 yields no removal, size = 2 gives mediocre removal and increasing the size past 4 gives no further improvement.

• Median blur test on the Cd spectrum image at 1.50mm with ksize=3:
  
• Adjusting the ksize to 1 yields no removal and increasing the ksize past 3 gives no further improvement.

• Both scipy.ndimage.filters.median_filter and cv2.medianBlur remove hot pixels from this test image with mostly the same result.

• Median filter test on the 20s dark frame with median_filter size 4:
  

• Median blur test on the 20s dark frame with ksize=3:
  

• Reducing the size and ksize below 4 and 3 respectively yields minimal noise reduction.
• If removing the hot pixels from the spectrum images works, is processing the dark frames a requirement?

• Median filter test on the Cd spectrum image at 6.50mm with size 4:
  

• Median blur test on the Cd spectrum image at 6.50mm with ksize=3:
  

• Testing code on a much brighter section of the spectrum (6.50mm) also shows hot pixel reduction

• hot_pixel.py.txt: Hot pixel removal test code - python - 20161013

Smoothing curves

• This figure shows a plot of the number of photoelectrons as a function of pixel position, the plots on the top are with the original values, the plots on the bottom are where the average is taken for the 10 adjacent values.
• Plot for Cd calibration for grating of 0.00mm:
  

• Calib_cd_0.00.txt: All of the local peaks and its location in the x axis for cad_plt_0.00 plot

• Calib_cd_6.50.txt: All of the local peaks and its location in the x axis for cad_plt_6.50 plot

• Calib_cd_10.0.txt: All of the local peaks and its location in the x axis for cad_plt_10.00 plot

• Plot for clad for grating of 10mm:
  

• plot_figures.zip: James' plots