A critical step to the reduction of HRS data is finding orders in the images. Typically, images of flat fields or of a bright object can be used to identify orders in the frame. The data must first be processed. Unfortunately, a truly perfect data set for the identification of the orders is rarely available as the response function across the CCD can widely vary.
Below we outline some tasks that can be used for identifying orders in HRS images and the steps that are used to process the data to make the identification possible.
Due to the changing response function across the CCD, a fiber flat image can show significant vignetting in both the vertical and horizontal direction. To remove the vignetting in the vertical direction, the normalize_image task can be used.
The normalize_image task fits a function to a fiber flat image after masking the image. Either a 1D or 2D function can be fit to the image, and the fitting function should be specified by the user. For the best performance, it is best to either apply an already existing order frame or to smooth the image and mask areas of low response.
Here is an example of the steps needed to normalize an image:
>>> from astropy.io import fits
>>> from astropy import modeling as mod
>>> from scipy import ndimage as nd
>>> from pyhrs import normalize_image
>>> hdu = fits.open('HFLAT.fits')
>>> image = nd.filters.maximum_filter(image, 10)
>>> mask = (image > 500)
>>> norm = normalize_image(image, mod.models.Legendre1D(10), mask=mask)
>>> norm[norm<10000] = 0
This will produce a ndarry where good areas all have the same value and bad areas will have values set to zero. This significantly simplifies the process of identifying orders in the image. However, orders at the very top of the image with little or no signal will still be difficult to detect.
The next step is to create an order frame. An order frame is defined as an image where each pixel associated with an order is identified and labeled with that order. To produce the order frame, an initial detection kernal (based on a user input) is convolved with a single column in the image. The first maximum identified is assocatied with the initial input order given by the user. To identify the full 2D shape of the order, all pixels above a certain threshold and connected are identified. These pixels are then given the value of the initial order. Once the order is identified, all pixels associated with this order are set to zero. The detection kernal is then updated based on the 1D shape of this order, the order is incremented, and the process is repeated until all orders are identified in the frame.
All of these steps are accomplished running the create_orderframe task. An example of running this task is the follow:
>>> norm[norm>500] = 500
>>> xc = int(norm.shape[0]/2)
>>> detect_kern = norm[30:110, xc]
>>> frame = create_orderframe(norm, 84, xc, detect_kern, y_start=30, y_limit=4050)
This will produce the order frame for the blue arm in medium resolution mode. It will identify all orders up to the limit of y-position of 4050. For the red arm, the first order is 53 for the medium resolution mode.