Welcome to the ccdproc documentation! Ccdproc is is an affiliated package for the AstroPy package for basic data reductions of CCD images. The ccdproc package provides many of the necessary tools for processing of ccd images built on a framework to provide error propogation and bad pixel tracking throughout the reduction process.

Documentation

The documentation for this package is here:

Installation

Requirements

Ccdproc has the following requirements:

• astropy v0.4 or later
• numpy
• scipy

One easy way to get these dependencies is to install a python distribution like anaconda.

Installing ccdproc

Using pip

To install ccdproc with pip, simply run:

pip install --no-deps ccdproc

Note

The --no-deps flag is optional, but highly recommended if you already have Numpy installed, since otherwise pip will sometimes try to “help” you by upgrading your Numpy installation, which may not always be desired.

Building from source

Obtaining the source packages

Source packages

The latest stable source package for ccdproc can be downloaded here.

Development repository

The latest development version of ccdproc can be cloned from github using this command:

git clone git://github.com/astropy/ccdproc.git

Building and Installing

To build ccdproc (from the root of the source tree):

python setup.py build

To install ccdproc (from the root of the source tree):

python setup.py install

Testing a source code build of ccdproc

The easiest way to test that your ccdproc built correctly (without installing ccdproc) is to run this from the root of the source tree:

python setup.py test

CCD Data reduction (ccdproc)

Introduction

Note

ccdproc works only with astropy version 0.4.0 or later.

The ccdproc package provides:

• An image class, CCDData, that includes an uncertainty for the data, units and methods for performing arithmetic with images including the propagation of uncertainties.
• A set of functions performing common CCD data reduction steps (e.g. dark subtraction, flat field correction) with a flexible mechanism for logging reduction steps in the image metadata.
• A class for combining and/or clipping images, Combiner, and associated functions.

Getting Started

Warning

ccdproc is still under active development. The API will almost certainly change.

In addition, testing of ccdproc on real data is currently very limited. Use with caution, and please report any errors you find at the GitHub repo for this project.

A CCDData object can be created from a numpy array (masked or not) or from a FITS file:

>>> import numpy as np
>>> from astropy import units as u
>>> import ccdproc
>>> image_1 = ccdproc.CCDData(np.ones((10, 10)), unit="adu")

An example of reading from a FITS file is image_2 = ccdproc.CCDData.read('my_image.fits', unit="electron") (the electron unit is defined as part of ccdproc).

The metadata of a CCDData object is a case-insensitive dictionary (though this may change in future versions).

The data is accessible either by indexing directly or through the data attribute:

>>> sub_image = image_1[:, 1:-3]  # a CCDData object
>>> sub_data =  image_1.data[:, 1:-3]  # a numpy array

See the documentation for CCDData for a complete list of attributes.

Most operations are performed by functions in ccdproc:

>>> dark = ccdproc.CCDData(np.random.normal(size=(10, 10)), unit="adu")
>>> dark_sub = ccdproc.subtract_dark(image_1, dark,
...                                  dark_exposure=30*u.second,
...                                  data_exposure=15*u.second,
...                                  scale=True)

Every function returns a copy of the data with the operation performed. If, for some reason, you wanted to modify the data in-place, do this:

>>> image_2 = ccdproc.subtract_dark(image_1, dark, dark_exposure=30*u.second, data_exposure=15*u.second, scale=True)

See the documentation for subtract_dark for more compact ways of providing exposure times.

Every function in ccdproc supports logging through the addition of information to the image metadata.

Logging can be simple – add a string to the metadata:

>>> image_2_gained = ccdproc.gain_correct(image_2, 1.5 * u.photon/u.adu, add_keyword='gain_corrected')

Logging can be more complicated – add several keyword/value pairs by passing a dictionary to add_keyword:

>>> my_log = {'gain_correct': 'Gain value was 1.5',
...           'calstat': 'G'}
>>> image_2_gained = ccdproc.gain_correct(image_2,
...                                       1.5 * u.photon/u.adu,
...                                       add_keyword=my_log)

The Keyword class provides a compromise between the simple and complicated cases for providing a single key/value pair:

>>> key = ccdproc.Keyword('gain_corrected', value='Yes')
>>> image_2_gained = ccdproc.gain_correct(image_2,
...                                       1.5 * u.photon/u.adu,
...                                       add_keyword=key)

Keyword also provides a convenient way to get a value from image metadata and specify its unit:

>>> image_2.header['gain']  = 1.5
>>> gain = ccdproc.Keyword('gain', unit=u.photon/u.adu)
>>> image_2_var = ccdproc.create_variance(image_2,
...                                       gain=gain.value_from(image_2.header),
...                                       readnoise=3.0 * u.photon)

You might wonder why there is a gain_correct at all, since the implemented gain correction simple multiplies by a constant. There are two things you get with gain_correct that you do not get with multiplication:

• Appropriate scaling of uncertainties.
• Units

The same advantages apply to operations that are more complex, like flat correction, in which one image is divided by another:

>>> flat = ccdproc.CCDData(np.random.normal(1.0, scale=0.1, size=(10, 10)),
...                        unit='adu')
>>> image_1_flat = ccdproc.flat_correct(image_1, flat)

In addition to doing the necessary division, flat_correct propagates uncertainties (if they are set).

Using ccdproc

Image class

Getting started

Getting data in

The tools in ccdproc accept only CCDData objects, a subclass of NDData.

Creating a CCDData object from any array-like data is easy:

>>> import numpy as np
>>> import ccdproc
>>> ccd = ccdproc.CCDData(np.arange(10), unit="adu")

Note that behind the scenes, NDData creates references to (not copies of) your data when possible, so modifying the data in ccd will modify the underlying data.

You are required to provide a unit for your data. The most frequently used units for these objects are likely to be adu, photon and electron, which can be set either by providing the string name of the unit (as in the example above) or from unit objects:

>>> from astropy import units as u
>>> ccd_photon = ccdproc.CCDData([1, 2, 3], unit=u.photon)
>>> ccd_electron = ccdproc.CCDData([1, 2, 3], unit="electron")

If you prefer not to use the unit functionality then use the special unit u.dimensionless_unscaled when you create your CCDData images:

>>> ccd_unitless = ccdproc.CCDData(np.zeros((10, 10)),
...                                unit=u.dimensionless_unscaled)

A CCDData object can also be initialized from a FITS file:

>>> ccd = ccdproc.CCDData.read('my_file.fits', unit="adu")  

If there is a unit in the FITS file (in the BUNIT keyword), that will be used, but a unit explicitly provided in read will override any unit in the FITS file.

There is no restriction at all on what the unit can be – any unit in astropy.units or that you create yourself will work.

Metadata

When initializing from a FITS file, the header property is initialized using the header of the FITS file. Metadata is optional, and can be provided by any dictionary or dict-like object:

>>> ccd_simple = ccdproc.CCDData(np.arange(10), unit="adu")
>>> my_meta = {'observer': 'Edwin Hubble', 'exposure': 30.0}
>>> ccd_simple.header = my_meta  # or use ccd_simple.meta = my_meta

Search of the metadata is case-insensitive:

>>> 'OBSERVER' in ccd_simple.header
True
>>> ccd_simple.header['ExPoSuRe']
30.0

Note, however, that internally all keywords are converted to lowercase.

Getting data out

A CCDData object behaves like a numpy array (masked if the CCDData mask is set) in expressions, and the underlying data (ignoring any mask) is accessed through data attribute:

>>> ccd_masked = ccdproc.CCDData([1, 2, 3], unit="adu", mask=[0, 0, 1])
>>> 2 * np.ones(3) * ccd_masked   # one return value will be masked
masked_array(data = [2.0 4.0 --],
             mask = [False False  True],
       fill_value = 1e+20)

>>> 2 * np.ones(3) * ccd_masked.data   # ignores the mask
array([ 2.,  4.,  6.])

You can force conversion to a numpy array with:

>>> np.asarray(ccd_masked)
array([1, 2, 3])
>>> np.ma.array(ccd_masked.data, mask=ccd_masked.mask)
masked_array(data = [1 2 --],
             mask = [False False  True],
       fill_value = 999999)

A method for converting a CCDData object to a FITS HDU list is also available. It converts the metadata to a FITS header:

>>> hdulist = ccd_masked.to_hdu()

You can also write directly to a FITS file:

>>> ccd_masked.write('my_image.fits')

Masks and flags

Although not required when a CCDData image is created you can also specify a mask and/or flags.

A mask is a boolean array the same size as the data in which a value of True indicates that a particular pixel should be masked, i.e. not be included in arithmetic operations or aggregation.

Flags are one or more additional arrays (of any type) whose shape matches the shape of the data. For more details on setting flags see astropy.nddata.NDData.

Uncertainty

Pixel-by-pixel uncertainty can be calculated for you:

>>> data = np.random.normal(size=(10, 10), loc=1.0, scale=0.1)
>>> ccd = ccdproc.CCDData(data, unit="electron")
>>> ccd_new = ccdproc.create_variance(ccd, readnoise=5 * u.electron)

See Gain correct and create variance for more details.

You can also set the uncertainty directly, either by creating a StdDevUncertainty object first:

>>> from astropy.nddata.nduncertainty import StdDevUncertainty
>>> uncertainty = 0.1 * ccd.data  # can be any array whose shape matches the data
>>> my_uncertainty = StdDevUncertainty(uncertainty)
>>> ccd.uncertainty = my_uncertainty

or by providing a ndarray with the same shape as the data:

>>> ccd.uncertainty = 0.1 * ccd.data
INFO: Array provided for uncertainty; assuming it is a StdDevUncertainty. [ccdproc.ccddata]

In this case the uncertainty is assumed to be StdDevUncertainty. Using StdDevUncertainty is required to enable error propagation in CCDData

If you want access to the underlying uncertainty use its .array attribute:

>>> ccd.uncertainty.array  
array(...)

Arithmetic with images

Methods are provided to perform arithmetic operations with a CCDData image and a number, an astropy Quantity (a number with units) or another CCDData image.

Using these methods propagates errors correctly (if the errors are uncorrelated), take care of any necessary unit conversions, and apply masks appropriately. Note that the metadata of the result is not set:

>>> result = ccd.multiply(0.2 * u.adu)
>>> uncertainty_ratio = result.uncertainty.array[0, 0]/ccd.uncertainty.array[0, 0]
>>> round(uncertainty_ratio, 5)
0.2
>>> result.unit
Unit("adu electron")
>>> result.header
CaseInsensitiveOrderedDict()

Note

In most cases you should use the functions described in Reduction toolbox to perform common operations like scaling by gain or doing dark or sky subtraction. Those functions try to construct a sensible header for the result and provide a mechanism for logging the action of the function in the header.

The arithmetic operators *, /, + and - are not overridden.

Combining images and generating masks from clipping

Note

No attempt has been made yet to optimize memory usage in Combiner. A copy is made, and a mask array constructed, for each input image.

The first step in combining a set of images is creating a Combiner instance:

>>> from astropy import units as u
>>> from ccdproc import CCDData, Combiner
>>> import numpy as np
>>> ccd1 = CCDData(np.random.normal(size=(10,10)),
...                unit=u.adu)
>>> ccd2 = ccd1.copy()
>>> ccd3 = ccd1.copy()
>>> combiner = Combiner([ccd1, ccd2, ccd3])

The combiner task really combines two things: generation of masks for individual images via several clipping techniques and combination of images.

Image masks/clipping

There are currently two methods of clipping. Neither affects the data directly; instead each constructs a mask that is applied when images are combined.

Masking done by clipping operations is combined with the image mask provided when the Combiner is created.

Min/max clipping

minmax_clipping masks all pixels above or below user-specified levels. For example, to mask all values above the value 0.1 and below the value -0.3:

>>> combiner.minmax_clipping(min_clip=-0.3, max_clip=0.1)

Either min_clip or max_clip can be omitted.

Sigma clipping

For each pixel of an image in the combiner, sigma_clipping masks the pixel if is more than a user-specified number of deviations from the central value of that pixel in the list of images.

The sigma_clipping method is very flexible: you can specify both the function for calculating the central value and the function for calculating the deviation. The default is to use the mean (ignoring any masked pixels) for the central value and the standard deviation (again ignoring any masked values) for the deviation.

You can mask pixels more than 5 standard deviations above or 2 standard deviations below the median with

>>> combiner.sigma_clipping(low_thresh=2, high_thresh=5, func=np.ma.median)

Note

Numpy masked median can be very slow in exactly the situation typically encountered in reducing ccd data: a cube of data in which one dimension (in the case the number of frames in the combiner) is much smaller than the number of pixels.

A much faster library for this case is bottleneck; a detailed example which uses bottleneck is at Faster medians using bottleneck.

Iterative clipping

To clip iteratively, continuing the clipping process until no more pixels are rejected, loop in the code calling the clipping method:

>>> old_n_masked = 0  # dummy value to make loop execute at least once
>>> new_n_masked = combiner.data_arr.mask.sum()
>>> while (new_n_masked > old_n_masked):
...     combiner.sigma_clipping(func=np.ma.median)
...     old_n_masked = new_n_masked
...     new_n_masked = combiner.data_arr.mask.sum()

Note that the default values for the high and low thresholds for rejection are 3 standard deviations.

Image combination

Image combination is straightforward; to combine by taking the average, excluding any pixels mapped by clipping:

>>> combined_average = combiner.average_combine()

Performing a median combination is also straightforward,

>>> combined_median = combiner.median_combine()  # can be slow, see below

The masked median function provided by numpy (and used by default in median_combine) can be very slow, so median_combine accepts an argument median_func for calculating the median instead. One fast alternative is provided by the bottleneck package; an example using it is at Faster medians using bottleneck.

With image scaling

In some circumstances it may be convenient to scale all images to some value before combining them. Do so by setting scaling:

>>> scaling_func = lambda arr: 1/np.ma.average(arr)
>>> combiner.scaling = scaling_func
>>> combined_average_scaled = combiner.average_combine()

This will normalize each image by its mean before combining (note that the underlying images are not scaled; scaling is only done as part of combining using average_combine or median_combine).

With image transformation

TBD

Reduction toolbox

Note

This is not intended to be an introduction to image reduction. While performing the steps presented here may be the correct way to reduce data in some cases, it is not correct in all cases.

Logging in ccdproc

All logging in ccdproc is done in the sense of recording the steps performed in image metadata. if you want to do logging in the python sense of the word please see those docs.

There are basically three logging options:

1. Implicit logging: No setup or keywords needed, each of the functions below adds a note to the metadata when it is performed.
2. Explicit logging: You can specify what information is added to the metadata using the add_keyword argument for any of the functions below.
3. No logging: If you prefer no logging be done you can “opt-out” by calling each function with add_keyword=None.

Gain correct and create variance

Uncertainty

An uncertainty can be calculated from your data with create_variance:

>>> from astropy import units as u
>>> import numpy as np
>>> import ccdproc
>>> img = np.random.normal(loc=10, scale=0.5, size=(100, 232))
>>> data = ccdproc.CCDData(img, unit=u.adu)
>>> data_with_variance = ccdproc.create_variance(data,
...                                              gain=1.5 * u.electron/u.adu,
...                                              readnoise=5 * u.electron)
>>> data_with_variance.header['exposure'] = 30.0  # for dark subtraction

The uncertainty, , at pixel with value is calculated as

where is the read noise. Gain is only necessary when the image units are different than the units of the read noise, and is used only to calculate the uncertainty. The data itself is not scaled by this function.

As with all of the functions in ccdproc, the input image is not modified.

In the example above the new image data_with_variance has its uncertainty set.

Gain

To apply a gain to an image, do:

>>> gain_corrected = ccdproc.gain_correct(data_with_variance, 1.5*u.electron/u.adu)

The result gain_corrected has its data and uncertainty scaled by the gain and its unit updated.

There are several ways to provide the gain, among them as an astropy.units.Quantity, as in the example above, as a ccdproc.Keyword. See to documentation for gain_correct for details.

Clean image

There are two ways to clean an image of cosmic rays. One is to use clipping to create a mask for a stack of images, as described in Image masks/clipping.

The other is to replace, in a single image, each pixel that is several standard deviations from a central value in a region surrounding that pixel. The methods below describe how to do that.

LACosmic

The lacosmic technique identifies cosmic rays by identifying pixels based on a variation of the Laplacian edge detection. The algorithm is an implementation of the code describe in van Dokkum (2001) [1].

Use this technique with cosmicray_lacosmic:

>>> cr_cleaned = ccdproc.cosmicray_lacosmic(gain_corrected, threshold,
...                                         thresh=5, mbox=11, rbox=11,
...                                         gbox=5)

median

Another cosmic ray cleaning algorithm available in ccdproc is cosmicray_median that is analogous to iraf.imred.crutil.crmedian. This technique can be used with ccdproc.cosmicray_median:

>>> cr_cleaned = ccdproc.cosmicray_median(gain_corrected, threshold,
...                                       mbox=11, rbox=11, gbox=5)

Although ccdproc provides functions for identifying outlying pixels and for calculating the deviation of the background you are free to provide your own error image instead.

There is one additional argument, gbox, that specifies the size of the box, centered on a outlying pixel, in which pixel should be grown. The argument rbox specifies the size of the box used to calculate a median value if values for bad pixels should be replaced.

Subtract overscan and trim images

Note

• Images reduced with ccdproc do NOT have to come from FITS files. The discussion below is intended to ease the transition from the indexing conventions used in FITS and IRAF to python indexing.
• No bounds checking is done when trimming arrays, so indexes that are too large are silently set to the upper bound of the array. This is because numpy, which provides the infrastructure for the arrays in ccdproc has this behavior.

Indexing: python and FITS

Overscan subtraction and image trimming are done with two separate functions. Both are straightforward to use once you are familiar with python’s rules for array indexing; both have arguments that allow you to specify the part of the image you want in the FITS standard way. The difference between python and FITS indexing is that python starts indexes at 0, FITS starts at 1, and the order of the indexes is switched (FITS follows the FORTRAN convention for array ordering, python follows the C convention).

The examples below include both python-centric versions and FITS-centric versions to help illustrate the differences between the two.

Consider an image from a FITS file in which NAXIS1=232 and NAXIS2=100, in which the last 32 columns along NAXIS1 are overscan.

In FITS parlance, the overscan is described by the region [201:232, 1:100].

If that image has been read into a python array img by astropy.io.fits then the overscan is img[0:100, 200:232] (or, more compactly img[:, 200:]), the starting value of the first index implicitly being zero, and the ending value for both indices implicitly the last index).

One aspect of python indexing may particularly surprising to newcomers: indexing goes up to but not including the end value. In img[0:100, 200:232] the end value of the first index is 99 and the second index is 231, both what you would expect given that python indexing starts at zero, not one.

Those transitioning from IRAF to ccdproc do not need to worry about this too much because the functions for overscan subtraction and image trimming both allow you to use the familiar BIASSEC and TRIMSEC conventions for specifying the overscan and region to be retained in a trim.

Overscan subtraction

To subtract the overscan in our image from a FITS file in which NAXIS1=232 and NAXIS2=100, in which the last 32 columns along NAXIS1 are overscan, use subtract_overscan:

>>> # python-style indexing first
>>> oscan_subtracted = ccdproc.subtract_overscan(cr_cleaned,
...                                              overscan=cr_cleaned[:, 200:],
...                                              overscan_axis=1)
>>> # FITS/IRAF-style indexing to accomplish the same thing
>>> oscan_subtracted = ccdproc.subtract_overscan(cr_cleaned,
...                                              fits_section='[201:232,1:100]',
...                                              overscan_axis=1)

Note well that the argument overscan_axis always follows the python convention for axis ordering. Since the order of the indexes in the fits_section get switched in the (internal) conversion to a python index, the overscan axis ends up being the second axis, which is numbered 1 in python zero-based numbering.

With the arguments in this example the overscan is averaged over the overscan columns (i.e. 2000 through 2031) and then subtracted row-by-row from the image. The median argument can be used to median combine instead.

This example is not very realistic: typically one wants to fit a low-order polynomial to the overscan region and subtract that fit:

>>> from astropy.modeling import models
>>> poly_model = models.Polynomial1D(1)  # one-term, i.e. constant
>>> oscan_subtracted = ccdproc.subtract_overscan(cr_cleaned,
...                                              overscan=cr_cleaned[:, 200:],
...                                              overscan_axis=1,
...                                              model=poly_model)

See the documentation for astropy.modeling.polynomial for more examples of the available models and for a description of creating your own model.

Trim an image

The overscan-subtracted image constructed above still contains the overscan portion. We are assuming came from a FITS file in which NAXIS1=2032 and NAXIS2=1000, in which the last 32 columns along NAXIS1 are overscan.

Trim it using trim_image,shown below in both python- style and FITS-style indexing:

>>> # FITS-style:
>>> trimmed = ccdproc.trim_image(oscan_subtracted,
...                              fits_section='[1:200, 1:100]')
>>> # python-style:
>>> trimmed = ccdproc.trim_image(oscan_subtracted[:, :200])

Note again that in python the order of indices is opposite that assumed in FITS format, that the last value in an index means “up to, but not including”, and that a missing value implies either first or last value.

Those familiar with python may wonder what the point of trim_image is; it looks like simply indexing oscan_subtracted would accomplish the same thing. The only additional thing trim_image does is to make a copy of the image before trimming it.

Note

By default, python automatically reduces array indices that extend beyond the actual length of the array to the actual length. In practice, this means you can supply an invalid shape for, e.g. trimming, and an error will not be raised. To make this concrete, ccdproc.trim_image(oscan_subtracted[:, :200000000]) will be treated as if you had put in the correct upper bound, 200.

Subtract bias and dark

Both of the functions below propagate the uncertainties in the science and calibration images if either or both is defined.

Assume in this section that you have created a master bias image called master_bias and a master dark image called master_dark that has been bias-subtracted so that it can be scaled by exposure time if necessary.

Subtract the bias with subtract_bias:

>>> fake_bias_data = np.random.normal(size=trimmed.shape)  # just for illustration
>>> master_bias = ccdproc.CCDData(fake_bias_data,
...                               unit=u.electron,
...                               mask=np.zeros(trimmed.shape))
>>> bias_subtracted = ccdproc.subtract_bias(trimmed, master_bias)

There are several ways you can specify the exposure times of the dark and science images; see subtract_dark for a full description.

In the example below we assume there is a keyword exposure in the metadata of the trimmed image and the master dark and that the units of the exposure are seconds (note that you can instead explicitly provide these times).

To perform the dark subtraction use subtract_dark:

>>> master_dark = master_bias.multiply(0.1)  # just for illustration
>>> master_dark.header['exposure'] = 15.0
>>> dark_subtracted = ccdproc.subtract_dark(bias_subtracted, master_dark,
...                                         exposure_time='exposure',
...                                         exposure_unit=u.second,
...                                         scale=True)

Note that scaling of the dark is not done by default; use scale=True to scale.

Correct flat

Given a flat frame called master_flat, use flat_correct to perform this calibration:

>>> fake_flat_data = np.random.normal(loc=1.0, scale=0.05, size=trimmed.shape)
>>> master_flat = ccdproc.CCDData(fake_flat_data, unit=u.electron)
>>> reduced_image = ccdproc.flat_correct(dark_subtracted, master_flat)

As with the additive calibrations, uncertainty is propagated in the division.

The flat is scaled by the mean of master_flat before dividing.

If desired, you can specify a minimum value the flat can have (e.g. to prevent division by zero). Any pixels in the flat whose value is less than min_value are replaced with min_value):

>>> reduced_image = ccdproc.flat_correct(dark_subtracted, master_flat,
...                                      min_value=0.9)

[1]	van Dokkum, P; 2001, “Cosmic-Ray Rejection by Laplacian Edge Detection”. The Publications of the Astronomical Society of the Pacific, Volume 113, Issue 789, pp. 1420-1427. doi: 10.1086/323894

Reduction examples

Mostly still TBD, hopefully filled in with examples from users. There is one example ipython notebook.

ccdproc Module

The ccdproc package is a collection of code that will be helpful in basic CCD processing. These steps will allow reduction of basic CCD data as either a stand-alone processing or as part of a pipeline.

Functions

background_deviation_box(data, bbox)	Determine the background deviation with a box size of bbox.
background_deviation_filter(data, bbox)	Determine the background deviation for each pixel from a box with size of bbox.
cosmicray_lacosmic(ccd[, error_image, ...])	Identify cosmic rays through the lacosmic technique.
cosmicray_median(ccd[, error_image, thresh, ...])	Identify cosmic rays through median technique.
create_deviation(ccd_data[, gain, ...])	Create a uncertainty frame.
flat_correct(ccd, flat[, min_value, add_keyword])	Correct the image for flat fielding.
gain_correct(ccd, gain[, gain_unit, add_keyword])	Correct the gain in the image.
rebin(ccd, newshape)	Rebin an array to have a new shape.
sigma_func(arr)	Robust method for calculating the deviation of an array.
subtract_bias(ccd, master[, add_keyword])	Subtract master bias from image.
subtract_dark(ccd, master[, dark_exposure, ...])	Subtract dark current from an image.
subtract_overscan(ccd[, overscan, ...])	Subtract the overscan region from an image.
test([package, test_path, args, plugins, ...])	Run the tests using py.test.
transform_image(ccd, transform_func[, ...])	Transform the image Using the function specified by transform_func, the transform will be applied to data, uncertainty, and mask in ccd.
trim_image(ccd[, fits_section, add_keyword])	Trim the image to the dimensions indicated.

Classes

CCDData(*args, **kwd)	A class describing basic CCD data The CCDData class is based on the NDData object and includes a data array, uncertainty frame, mask frame, meta data, units, and WCS information for a single CCD image.
Combiner(ccd_list)	A class for combining CCDData objects.
Keyword(name[, unit, value])

Class Inheritance Diagram