LUCY

←Prev
KAPPA –- Kernel Application Package
Next→
TOC ↑

LUCY

Performs a Richardson-Lucy deconvolution of a one- or two-dimensional array

Description:

This application deconvolves the supplied one- or two-dimensional array using the Richardson-Lucy (R-L) algorithm. It takes an array holding observed data and another holding a Point-Spread Function (PSF) as input and produces an output array with higher resolution. The algorithm is iterative, each iteration producing a new estimate of the restored array which (usually) fits the observed data more closely than the previous estimate (in the sense that simulated data generated from the restored array is closer to the observed data). The closeness of the fit is indicated after each iteration by a normalised

$χ^{2}$ value (i.e. the

$χ^{2}$ per pixel). The algorithm terminates when the normalised

$χ^{2}$ given by Parameter AIM is reached, or the maximum number of iterations given by Parameter NITER have been performed. The current estimate of the restored array is then written to the output NDF .

Before the first iteration, the restored array is initialised either to the array given by Parameter START, or, if no array is given, to the difference between the mean value in the input data array and the mean value in the background (specified by Parameters BACK and BACKVAL). Simulated data are then created from this trial array by smoothing it with the supplied PSF, and then adding the background on. The $χ^{2}$ value describing the deviation of this simulated data from the observed data are then found and displayed. If the required $χ^{2}$ is not reached by this simulated data, the first iteration commences, which consists of creating a new version of the restored array and then creating new simulated data from this new restored array (the corresponding $χ^{2}$ value is displayed). Repeated iterations are performed until the required $χ^{2}$ is reached, or the iteration limit is reached. The new version of the restored array is created as follows.

(1): A correction factor is found for each data value. This is the ratio of the observed data value to the simulated data value. An option exists to use the Snyder modification as used by the LUCY program in the STSDAS package within IRAF. With this option selected, the variance of the observed data value is added to both the numerator and the denominator when finding the correction factors.
(2): These correction factors are mapped into an array by smoothing the array of correction factors with the transposed PSF.
(3): The current version of the restored array is multiplied by this correction factor array to produce the new version of the restored array.

For further background to the algorithm, see L.B. Lucy, Astron.J. 1974, Vol 79, No. 6.

Usage:

lucy in psf out [aim]

Parameters:

AIM = _REAL (Read)

The

$χ^{2}$ value at which the algorithm should terminate. Smaller values of AIM will result in higher apparent resolution in the output array but will also cause noise in the observed data to be interpreted as real structure. Small values will require larger number of iterations, so NITER may need to be given a larger value. Very-small values may be completely un-achievable, indicated by

$χ^{2}$ not decreasing (or sometimes increasing) between iterations. Larger values will result in smoother output arrays with less noise. [1.0]

BACK = NDF (Read)

An NDF holding the background value for each observed data value. If a null value is supplied, a constant background value given by Parameter BACKVAL is used. [!]

BACKVAL = _REAL (Read)

The constant background value to use if BACK is given a null value. [0.0]

CHIFAC = _REAL (Read)

The normalised

$χ^{2}$ value which is used to determine if the algorithm should terminate is defined as follows:

$χ^{2} = \frac{1}{N} . \sum \frac{{(d - s)}^{2}}{(C H I F A C . s - σ^{2})}$

where the sum is taken over the entire input array (excluding the margins used to pad the input array), n is the number of values summed, $d$ is the observed data value, $s$ is the simulated data value based on the current version of the restored array, $σ^{2}$ is the variance of the error associated with $d$ , and $C H I F A C$ is the value of Parameter CHIFAC. Using 0 for CHIFAC results in the standard expression for $χ^{2}$ . However, the algorithm sometimes has difficulty fitting bright features and so may not reach the required normalised $χ^{2}$ value. Setting CHIFAC to 1 (as is done by the LUCY program in the STSDAS package within IRAF) causes larger data values to be given less weight in the $χ^{2}$ calculation, and so encourages lower $χ^{2}$ values. [1.0]

IN= NDF (Read)

The input NDF containing the observed data.

NITER = _INTEGER (Read)

The maximum number of iterations to perform. [50]

OUT = NDF (Write)

The restored output array. The background specified by Parameters BACK and BACKVAL will have been removed from this array. The output is the same size as the input. There is no VARIANCE component in the output, but any QUALITY values are propagated from the input to the output.

PSF = NDF (Read)

An NDF holding an estimate of the Point-Spread Function (PSF) of the input array. This could, for instance, be produced using the Kappa application PSF. There should be no bad pixels in the PSF otherwise an error will be reported. The PSF can be centred anywhere within the array, but the location of the centre must be specified using Parameters XCENTRE and YCENTRE. The PSF is assumed to have the value zero outside the supplied NDF.

SIGMA = _REAL (Read)

The standard deviation of the noise in the observed data. This is only used if Parameter VARIANCE is given the value FALSE. If a null (!) value is supplied, the value used is an estimate of the noise based on the difference between adjacent pixel values in the observed data. [!]

START = NDF (Read)

An NDF containing an initial guess at the restored array. This could, for instance, be the output from a previous run of LUCY, in which case the deconvolution would continue from the point it had previously reached. If a null value is given, then the restored array is initialised to a constant value equal to the difference between the mean observed data value and the mean background value. [!]

SNYDER = _LOGICAL (Read)

If TRUE then the variance of the observed data sample is added to both the numerator and denominator when evaluating the correction factor for each data sample. This is the modified form of the R-L algorithm used by the LUCY program in the STSDAS package within IRAF. [TRUE]

THRESH = _REAL (Read)

The fraction of the PSF peak amplitude at which the extents of the PSF are determined. These extents are used to determine the size of the margins used to pad the supplied input array. Lower values of THRESH will result in larger margins being used. THRESH must be positive and less than 0.5. [0.0625]

TITLE = LITERAL (Read)

A title for the output NDF. A null (!) value means using the title of the input NDF. [!]

VARIANCE = _LOGICAL (Read)

If TRUE, then the variance of each input data sample will be obtained from the VARIANCE component of the input NDF. An error is reported if this option is selected and the NDF has no VARIANCE component. If FALSE, then a constant variance equal to the square of the value given for Parameter SIGMA is used for all data samples. If a null (!) value is supplied, the value used is TRUE if the input NDF has a VARIANCE component, and FALSE otherwise. [!]

WLIM = _REAL (Read)

If the input array contains bad pixels, then this parameter may be used to determine the number of good data values which must contribute to an output pixel before a valid value is stored in the restored array. It can be used, for example, to prevent output pixels from being generated in regions where there are relatively few good data values to contribute to the restored result. It can also be used to ‘fill in’ small areas (i.e. smaller than the PSF) of bad pixels.

The numerical value given for WLIM specifies the minimum total weight associated with the good pixels in a smoothing box required to generate a good output pixel (weights for each pixel are defined by the normalised PSF). If this specified minimum weight is not present, then a bad output pixel will result, otherwise a smoothed output value will be calculated. The value of this parameter should lie between 0.0 and 1.0. WLIM=0 causes a good output value to be created even if there is only one good input value, whereas WLIM=1 causes a good output value to be created only if all input values are good. Values less than 0.5 will tend to reduce the number of bad pixels, whereas values larger than 0.5 will tend to increase the number of bad pixels.

This threshold is applied each time a smoothing operation is performed. Many smoothing operations are typically performed in a run of LUCY, and if WLIM is larger than 0.5 the effects of bad pixels will propagate further through the array at each iteration. After several iterations this could result in there being no good data left. An error is reported if this happens. [0.001]

XCENTRE = _INTEGER (Read)

The x pixel index of the centre of the PSF within the supplied PSF array. If a null (!) value is supplied, the value used is the middle pixel (rounded down if there are an even number of pixels per line). [!]

YCENTRE = _INTEGER (Read)

The y pixel index of the centre of the PSF within the supplied PSF array. If a null (!) value is supplied, the value used is the middle line (rounded down if there are an even number of lines). [!]

Examples:

lucy m51 star m51_hires

This example deconvolves the array in the NDF called m51, putting the resulting array in the NDF called m51_hires. The PSF is defined by the array in NDF star (the centre of the PSF is assumed to be at the central pixel). The deconvolution terminates when a normalised chi-squared value of 1.0 is reached.

lucy m51 star m51_hires 0.5 niter=60

This example performs the same function as the previous example, except that the deconvolution terminates when a normalised chi-squared value of 0.5 is reached, giving higher apparent resolution at the expense of extra spurious noise-based structure. The maximum number of iterations is increased to 60 to give the algorithm greater opportunity to reach the reduced chi-squared value.

lucy m51 star m51_hires2 0.1 start=m51_hires

This example continues the deconvolution started by the previous example in order to achieve a normalised chi-squared of 0.1. The output array from the previous example is used to initialise the restored array.

Notes:

The convolutions required by the R-L algorithm are performed by the multiplication of Fourier transforms. The supplied input array is extended by a margin along each edge to avoid problems of wrap-around between opposite edges of the array. The width of this margin is about equal to the width of the significant part of the PSF (as determined by Parameter THRESH). The application displays the width of these margins. The margins are filled by replicating the edge pixels from the supplied input NDFs.
The R-L algorithm works best for arrays which have zero background. Non-zero backgrounds cause dark rings to appear around bright, compact sources. To avoid this a background array should be created before running LUCY and assigned to the Parameter BACK. The SEGMENT and SURFIT applications within Kappa can be used to create such a background array.

Related Applications

KAPPA: FOURIER, MEM2D, WIENER.

Implementation Status:

This routine correctly processes the AXIS, DATA, QUALITY, VARIANCE, LABEL, TITLE, UNITS, WCS, and HISTORY components of the input NDF and propagates all extensions.
Processing of bad pixels and automatic quality masking are supported.
All non-complex numeric data types can be handled. Arithmetic is performed using single-precision floating point.

←Prev
KAPPA –- Kernel Application Package
Next→
TOC ↑