| | 19 | This new technique works roughly as well on o5579g0223o.XY63 used in the earlier discussion. Below is a region from o5579g0223o.XY63 with a small amount of smoothing to enhance the detections. Red boxes mark (unculled) detections using the default technique for culling on the raw image; Blue boxes mark (unculled) detections using the new culling technique with the smoothed image. Clearly, this technique is filtering out many of the insignificant detections from the poorly subtracted background model. |
| | 20 | |
| | 21 | [[Image(o5579g0223o.XY63.jpg))]] |
| | 22 | |
| | 23 | Unfortunately, the technique is not perfect. The two images below show a different image (o5353g0100o.XY20) with somewhat similar features. In this case, the excess flux comes from dark glows in the relatively poor corners (near the area masked for bad CTE and dark current). In this example, these detections are not efficiently culled from these images with the new technique. (The left image is unsmoothed; the right has a 2.0 sigma Gaussian smoothing applied). The difficulty in this case is that these detections land on very low significance fluctuations in the background -- according to the significance image, they are significant (if barely), and the intervening cols are below the minimum threshold. Formally, these are significant. |
| | 24 | |
| | 25 | || [[Image(o5353g0100o.XY20.raw.jpg)]] || [[Image(o5353g0100o.XY20.smooth.jpg)]] || |
| | 26 | |
| | 27 | The final image below shows an example of some of the other ugly features in these images: star glints and burntool residuals are the most obvious. |
| | 28 | |
| | 29 | [[Image(o5353g0100o.XY20.crud.jpg)]] |
