UK Solar System Data Centre
The level 1 data set is dominated by light scattered from the f-conona.
But on top of this there are signals from CMEs, from the suns streamer belts, from stars and optical effects in the camera itself. To estimate the f-conona, people often take a simple pixel by pixel minima of a set of images. Following discusions with Andy Buffington we have attempted a more analytical approach where we take a running window of either 1 or 11 days and within that window select the lowest 25% of the data and form an average of that on a pixel by pixel basis - this is then subtracted from the original image to make the L2_1_25 and L2_11_25 data sets.
The following images are scaled with a dynamic range 100 times less than that of the f-corona image above. The first image effecticely excludes the streamer belt in hi-1, the second shows the streamer belt.
Another way of looking at this subtraction process is to look at the time sequence of a single pixel; the following plots show 9 such sequences arranged in a grid across the hi-1 field of view both before and after the subtaction.
For hi-2 the stars do not drift far enough in a single day to be optimally processed by the above algorithm - I would strongly encourage the use of the longer backgrounds for these cameras.
Periodically the hi-1B camera jumps in pointing direction - thought to be due to dust imacts since the camera is mounted facing the ram direction of the spacecraft. If the offset in the image sequence is more that one pixel then the images are corrected to allow for the pointing change before the f-corona is computed.