MOPEX Online Manual

Single Frame Outlier Rejection

Figure 1: Cartoon of single-frame outlier detection.

The single frame radhit detection represents spatial filtering of input images. It is implemented in the module Detect Radhit. It is a variant of image segmentation and is based on the idea that when a relatively low detection threshold is applied, for each bright point source a large number of pixels will be detected above the threshold. Bright detections with small areas are, therefore, classified as radhits. It should be applied to background subtracted images.

It has three input parameters: Segmentation Threshold, Radhit Threshold, and Detection Max Area. It performs image segmentation based on a low threshold Segmentation Threshold and then applies two conditions to weed out point sources. The first condition is that the total area in pixels should be no greater than a user specified Detection Max Area. The second condition is that at least one pixel in the cluster should be greater than another user specified Radhit Threshold, which should be set much higher than Segmentation Threshold.

Figure 2 illustrates the above algorithm. The three objects depicted there are a bright point source, a faint point source, and a radhit. After segmentation with a low Segmentation_Threshold is performed, the bright point source creates a cluster of pixels with the number of pixels greater than Detection Max Area. Then the second criterion of having at least one pixel greater than Radhit Threshold weeds out the faint point source.

Figure 2: An illustration of how Detect Radhit selects a radhit over two point sources.