Improving Hot Pixel Detection Algorithm

• Set out to create an improved algorithm for detecting hot pixels,
• Set out to create a 3 by 3 pixel square which traces across the whole CCD chip, and looked at the 8 pixels surrounding the central one, and taking the median (mode could be an issue is no repeating values exist). This value is then compared to the central pixel, and if it is sufficiently larger than its surrounding its x,y and intensity values are noted.
• Note, using this method pixels on the boundary of the image cannot be checked, as these do not have 8 nearest neighbours, so are not included using this method.
• Also, using this method, a threshold of sorts still needs to be applied, when comparing the central pixel to the surrounding median.
• Using this method, for similar thresholds for this and the previous method, it was found that this gave a slightly smaller value than was found previously (although of the same order of magnitude). Although it is worth noting our thresholds for the new method vary slightly with the median, although they remain relatively similar.

Overlaying found hot pixels over original image

• We overlayed our found hot pixels over the original image by setting the threshold as . This was done because usually the median threshold was around 160, whereas a typical hot pixel was more than 10000.
• This is what we found. The pink points are the pixels detected.

• If zoomed in we can see that the hot pixels were detected:

• However, some hot pixels were not detected if the surrounding pixels were bled into. This is because the median would be a high value as compared to the actual background, and therefore the threshold there was too high.

• However, we noticed that at low exposures the pixels did not bleed. As we will be working at lower exposures, this should hopefully not prove to be too large an issue.

• If the threshold is set to 3 standard deviations of the median we get the following overlay. Note that due to random effects in the CCD, around 4500 hot pixels are expected to be seen due to statistical fluctuations ( 32761 hot pixels found).

• Zoomed in it looks like the following:

• It appears that it selected pixels that do not appear to be hot pixels.
• At 5 standard deviations, this is what the overlay looks like:

• 26860 hot pixels were found for 5 standard deviations.
• At 10 standard deviations 19200 hot pixels were found and this is what the overlay looked like:

• At 20 standard deviations 12696 hot pixels were found. The overlay is shown below:

'MIP' Peak Investigation

• As seen last week, we observed that in the intensity of the hot pixels distribution (histogram) there was a unusually large peak present that did not fit in with the background. It was theorised that this could be due to a MIP (most ionising particle), and we investigated this here.
• We isolated the peak from the previous image, and found the x and y positions of the hot pixels and overlayed this with the original image, the results of which can be seen below (for both upwards and angled 1 hour exposures).

• From this it can be seen that this is not from MIP's, as these would be randomly distributed over the chip, and instead they can be seen clustered around the edges of the chip, and this is perhaps due to electronic noise or other affects.