Susan R. Stolovy University of Arizona, Steward Observatory, Tucson, AZ 85721, USA, e-mail: firstname.lastname@example.org
NICMOS, Image Restoration, OMC-1, deconvolution
The NICMOS cameras on HST at longer wavelengths allow well-sampled diffraction limited imaging with a very high dynamic range which is achieved using multiple non-destructive readouts of the arrays. During the check-out phase, soon after NICMOS was installed, images were taken of the OMC-1 region using the F212N and F215N narrow-band filters to isolate the 2.12 m molecular hydrogen line. These images formed part of the NICMOS early release (ERO) data and their scientific content has been described in Stolovy et al. (1998). The high dynamic range, good sampling, excellent signal-to-noise ratio and stable point-spread function of these NICMOS images suggests that image restoration methods may be able to significantly enhance them. It was hoped that such processing, which enhances contrast and resolution, could permit better study of the complex morphology of the highly structured environment close to the BN object. A particular problem with NICMOS images is that the point-spread functions exhibit rich structure as a result of the complex obscuration pattern in the pupil coming from both the HST OTA and the cold mask in the NICMOS dewar. Also, for very bright sources such as BN, the diffraction spikes can cover the entire array. The second aim of this work was to try to suppress PSF structure in these images so allow the underlying image to be better seen. This work is currently in progress and the results presented must be regarded as highly provisional.
The standard NICMOS data reduction was performed using a customized version of the CALNICA software developed at STScI. There were seven images at each position on the target and these were combined to produce five processed images in each filter at different pointings and orientations with significant overlaps. The five images through the F212N filter form the basis of the image restoration attempts. This filter includes both the molecular hydrogen emission and the continuum and hence shows the most diffuse structure. It also allows us to simultaneously look for sharpening of H2 features as they tend to be spatially anti-correlated with the continuum emission (see Figure 3 of Stolovy et. al 1998). The first phase of the image restoration was to obtain an adequate PSF. For the central frame which includes the very bright BN object a PSF from the Tiny Tim simulator (Krist & Hook 1997) was used. The most recent version of this code contains residual aberrations measured by phase retrieval as well facilities for defocus and compensation for cold-mask shifts in the pupil. For the outer frames a PSF extracted from a star in a relatively uncrowded region was extracted and scaled after suitable background subtraction. The empirical PSF matched the central structure somewhat better than the Tiny Tim one but had inadequate signal in the outer parts to model the full PSF of the BN object and small errors in the background subtraction led to significant artifacts. Standard image restoration in crowded fields with strong, varying, background lead to artifacts around bright stars. These are typically circular dark rings centred on stars and were found to be very pronounced in this NICMOS data. Instead of taking a direct approach, a two channel variant of the Richardson-Lucy method was employed (Lucy 1994; Hook & Lucy 1994, Hook et al. 1994). This method used a list of positions of known point-sources which have fixed positions in combination with a background image which is constrained to be smooth on a user-specified scale. As iterations proceed the intensities of the points model the stars and the background is modelled in the other channel. This second background channel is constrained to be smooth by the addition of an entropic term in the objective function being maximised. In the original version of this method (known as PLUCY) the points were represented as delta functions in an image and hence had to be centred in pixels. In a newer enhancement (CPLUCY) this constraint is removed and the stars can be assigned positions freely at a subpixel level. This method was found to effectively remove the outer parts of stars images although minor residuals are visible close to the centres.
In the very difficult case of the bright BN object the diffraction rings are largely cleaned up but there are residuals further out in the spider spikes. The two-channel restoration of the central frame is shown in Figure 1. A bright point-source, approximately 0.9 arcsecs to the NW of the BN object becomes apparent when the blobs in the diffraction pattern very close to the centre are suppressed. The exact nature of the point is unclear but it does not appear to be a ghost image from the filter.
The application of image restoration allows many fine-scale structures in these images to be more clearly seen and less affected by the PSFs from bright stars. This has allowed the region close to the BN object to be better studied than before and a point source 0.9 arcsecs to the NW has been noted but its exact nature is unclear. It may be a bright blob of dust and gas reflecting light from BN or it could be a previously undiscovered star. There is also intriguing faint asymmetric diffuse structure close to BN. The restorations are limited by the accuracy of the PSFs available. Tiny Tim does a good but not perfect job in this case, presumably because of an incomplete knowledge of the geometrical parameters of the NICMOS camera - residual aberrations, cold-mask shifts and so forth. Empirical PSFs taken from the images do not have enough signal to map the outer spider diffraction features which are visible around the BN object.
This deconvolution technique was successful in enhancing details in the extended molecular hydrogen ``fingers" and ``bullets" (see Stolovy et al. 1998). The southern, illuminated edge of the continuum ``Crescent" feature north of BN is also sharpened.
This work is in progress and will be extended to determining the best way to generally deconvolve narrow-band emission line images from NICMOS. This effort is complicated by the fact that the line and continuum filters have different PSFs. In this case, the F215N image would normally be subtracted from the F212N image to produce the molecular hydrogen image.