Wednesday, January 09, 2008

Interference Removal from Time-Frequency Data

We start with a flat time-frequency data, much like shown in the image on the left. This is VLA data (one baseline) observed at 325 MHz. The arrows indicate high-level and low-level interference in the data.

The most striking quality of interference, is that it is mostly confined to high 'spatial' frequencies. Here, I refer to spatial grey-level variation frequencies.

There are two approaches to remove / isolate such an interference:
  1. To take an FT of the image itself.
  2. Remove the high-frequency features (those belonging to fast-changing features.
  3. For all pixels, use a large (21x21 pixel) filter to compute mean, median and rms of grey levels.
  4. If the pixel value is greater than (mean+5*sigma) or (median + 5 sigma), then one can safely substitute median for the pixel value.
  5. One could compute rms over the entire image to avoid being dominated by local features.


Anonymous said...

I want not agree on it. I over warm-hearted post. Particularly the designation attracted me to review the whole story.

Anonymous said...

Amiable post and this mail helped me alot in my college assignement. Gratefulness you seeking your information.