% @(#)deconvfluc.hlq 17.1.1.1 (ESO-ECF) 01/25/02 17:15:05 %++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ %.COPYRIGHT (c) 1992 European Southern Observatory %.IDENT deconvfluc.hlq %.AUTHOR RNH, ECF/ESO %.KEYWORDS MIDAS, help files, DECONVOLVE/FLUCY %.PURPOSE On-line help file for the command: DECONVOLVE/FLUCY %.VERSION 1.0 30-SEP-1992 : Creation, RNH %---------------------------------------------------------------- \se SECTION./FLUCY \es\co DECONVOLVE/FLUCY 30-SEP-1992, RNH \oc\su DECONVOLVE/FLUCY images PSFs niter decon coaddr [flags] [xsub,ysub] [Firstim] Deconvolve one or more images using FFT based Lucy algorithm \us\pu Purpose: Deconvolve one or more images using FFT based Lucy algorithm \up\sy Syntax: DECONVOLVE/FLUCY in_spec PSFs niter decon coaddr [flags] [xsub,ysub] [Firstim] \ys\pa in_spec = one or more image frames separated by a comma;\\ the images must be well-aligned with each other and must have same size as well as dimensions which are powers of two. \ap\pa PSFs = point-spread-function images to match the input images;\\ the PSFs must be the same size and shape as the data images.\\ They must be non-negative everywhere and must be normalised so that the total intensity for each image is 1.0.\\ The peak of the PSFs should be positioned centrally in the frame and the PSFs should be well aligned with each other. \ap\pa niter = number of iterations to be applied \ap\pa decon = name of output deconvolved image;\\ this frame will be the result of applying the Lucy-Richardson restoration algorithm to the input images.\\ If there are multiple images the information from all of them is combined.\\ If there is only one image and one PSF this result is the standard Lucy-Richardson restored image. \ap\pa coaddr = root name for output co-added images\\ The co-added output images are the results of convolving the deconvolved image with each of the PSFs in turn.\\ They are named by appending _1, _2 etc to the string supplied here \ap\pa Flags = control switches, four Y/N flags; defaulted to YYNN\\ (1) verbose - if "Y" lots of messages explaining what is happening during the restoration are displayed. If "N" the program keeps quiet until all is done. \\ (2) accel - whether or not the accelerated version of the algorithm is used. \\ If this is turned on the correction factor image which is applied at each iteration is multiplied by a factor which maximises the increase in likelihood (and hence the improvement in the restoration) within the limits imposed by the non-negativity requirement. \\ This normally leads to quicker restoration and is recommended whether or not you have understood the last sentence. \\ (3) Whether or not there is a first estimate image. If so the restoration starts from this point, if not it starts from a constant image with the same total flux as the data. This option is mainly useful if you want to continue a restoration to more iterations without having to start again. \\ (4) Whether or not the restoration uses sub-sampling. It is often advantageous to restore onto a more finely sampled grid than the original data, particularly if the latter is not well sampled. For more details see the notes below. \ap\pa xsub,ysub = optional X and Y sub-sampling factors. These values are only needed if the subsampling flag above has been set to Y. \ap\pa Firstim = the name of the starting image if there is one. If the flag has been set to state that there is a starting image this one will be read and used as the initial estimate image. If not it is ignored. \ap\see See also: DECONVOLVE/IMAGE, COADD/IMAGE \ees\no Note: This program is invoked using either the DECONVOLVE/FLUCY command or the COADD/IMAGE command. In both cases the same code is executed but the two names reflect the two uses of this command: \\ Image restoration to enhance resolution and the contrast of fine scale features and the co-adding of images with different point-spread-functions, eg, CCD frames taken in differing seeing conditions. \\ Both operations are based on the image restoration scheme devised by Richardson and Lucy and the generalisation of the method for image co-addition was also suggested by Leon Lucy. \\ This command always produces both a deconvolved image and the result of convolving this with the PSFs of the input frames. \\ If the program is being used simply to do a normal restoration then the former will be the desired output and the latter may be disregarded (however it will be found to be very similar to the input with the pixel-to-pixel small-scale noise removed and hence may have its uses). \\ For simple restorations 40 iterations will typically be a good compromise between enhanced sharpness and obtrusive artefacts, although this number depends on the signal-to-noise of the data and other factors which make experimentation essential. \\ In the case of image co-addition the optimal result is normally the output co-added image which corresponds to the sharpest input PSF. \\ E.g., if you have three input images taken in 0.3,0.8 and 1.2 arcsec seeing then the optimal result will be the one matching the 0.3arsec PSF. \\ The deconvolved image may have smaller star images but it will also suffer from the unavoidable artefacts present in restorations. \\ For co-addition work it is best to go to quite high numbers of iterations to try to get close to the maximum likelihood result. \\ Sub-sampling may be done if the PSF can be obtained on a finer grid than the data itself. In this case the data should be expanded by pixel replication (not re-sampling or interpolation) so that the size and number of pixels matches that of the sub-sampled PSFs. \\ Then the restoration/co-addition may be performed in the normal way except that the sub-sampling flag should be set to Y and the appropriate values for the X and Y sub-sampling factors also given. \on\exs Examples: \ex DECONVOLVE/FLUCY galaxy psf 10 gal_10 qqq Do a standard Lucy-Richardson restoration of the image `galaxy.bdf' using the PSF `psf.bdf'. \\ 10 iterations are applied (accelerated as that is the default) and the result written to `gal_10.bdf'. \\ The result of convolving `gal_10.bdf' with `psf.bdf' is written to the scratch file `qqq.bdf' and will probably be disgarded. \xe \ex DECONVOLVE/FLUCY gal1,gal2 psf1,psf2 100 gal_a100 gal_ca Co-add the images `gal1.bdf' and `gal2.bdf' using the corresponding PSFs `psf1.bdf' and `psf2.bdf'. \\ 100 iterations (accelerated) will be performed. The deconvolved result will be written to `gal_a100.bdf' and the co-added images to `gal_ca_1.bdf' and `gal_ca_2.bdf'. \xe \ex DECONVOLVE/FLUCY galaxy psf 10 gal_20 qqq2 YYYN ? gal_10 Continue the restoration in the first example to another 10 iterations. \\ This shows the use of the `first estimate' flag which is set to "Y" and the input first estimate image name is given as `gal_10 'on the command line. \\ The result will be written to `gal_20.bdf'. The normal "verbose" and "accel" flags are retained. \xe \ex DECONVOLVE/FLUCY biggal psfsub 10 gal_s40 qqq YYNY 2,2 This is an example of using the sub-sampling restoration option. The input `biggal.bdf' is the same frame as above except with the pixels duplicated in both X and Y. \\ The PSF `psfsub.bdf' is an image of the same size and with the same pixel grid containing a sub-sampled PSF. The result will be `gal_s40.bdf', also on the finer pixel grid. The "subsample" flag has been set and the sub-sampling factors in both X and Y are given on the command line. \xe\sxe