BL: Correct an Image for Baseline Subtraction Noise



NOTE: This is an old Lick Observatory data-taking system command that is no longer valid (unless you have to reduce data taken before 1990). It is not the same as overscan subtraction.

BL removes the noise introduced through the digital baseline-restoration procedure used by the data collection program. The last column of the CCD, called the 'baseline column' is not illuminated during an exposure. Each pixel in this column determines the zero-level for the corresponding row in the rest of the chip, which is subtracted from each pixel in that row as the image is read off the chip into the data-taking system. NOTE THAT THIS IS THE PROCEDURE IMPLEMENTED AT LICK OBSERVATORY, BUT NOT MOST OTHER PLACES! Lick subtracts a row by row baseline value before storing the image at the telescope. If this is not done at your site, you should not use the BL command! Write a procedure to do the appropriate bias and/or baseline subtraction using the information supplied by your data taking system.

The baseline measurement is rather noisy, so when the zero-levels are subtracted, significant variations in level from one row to another are introduced. The BL program corrects for this by fitting a straight line by least-squares to the values in the baseline column. The original baseline measurements are then added back into the image data and the mean baseline values evaluated from the linear least-squares fit are subtracted. The best fit baseline is stored in the baseline column, unless the keyword KEEP is specified, in which case the original baseline is left in the baseline column.

The baseline column can sometimes exhibit several jumps in level along its length. BL detects these jumps, fitting a series of linear functions to the baseline column between these jumps. To detect the jumps, thus fitting the entire column with more than one linear function, use the JUMP keyword.

The BL procedure should be applied to every raw image before analysis or processing.