fixpix: Fix bad pixels by linear interpolation from nearby pixels

Package: proto

Usage

fixpix images masks

Parameters

images: List of two dimensional images to be "fixed" (modified) by linear interpolation.

masks: List of bad pixel masks, images, or files (collectively called masks) identifying the bad pixels. The list of masks must either match the list of input images in number or a single mask may be specified to apply to all images. The special name "BPM" may be specified to select a mask specified by the header keyword "BPM" in the input image. The possible mask formats are given in the DESCRIPTION section.

linterp = "INDEF", cinterp = "INDEF": Normally interpolation is performed across the narrowest dimension spanning the bad pixels with interpolation along image lines if the two dimensions are equal. However specific values in the mask may be used to identify the desired interpolation direction. The value in the mask specifying line interpolation is given by the linterp parameter and the value specifying column interpolation is given by the cinterp parameter. Any values which are do not match one of these parameters results in interpolation along the narrowest dimension. Note that a text file mask always has 2 for pixels with narrow dimension along lines and 3 for pixels with narrow dimension along columns.

verbose = no: If this parameter is set to yes a line identifying each image and associated mask is printed. If the pixels parameter is set then a list of the pixels modified is also printed.

pixels = no: List the pixels modified? This is only done if this parameters and the verbose parameter are set.

Description

Pixels in a list of images identified by bad pixel masks, images, or text files (collectively called masks here) are replaced by linear interpolation along lines or columns using the nearest good pixels. The list of input images, specified by the images parameter, are matched with a list of masks, specified by the masks parameter. The list of masks must match the list of input images or a single mask name may be given to apply to all input images. The special mask name "BPM" may be used to select a mask name given in the input image header under the keyword "BPM".

There are three types of masks which may be used. The preferred type is a bad pixel mask given as a "pixel list" image. Pixel list images have the extension ".pl" signifying a special compact file of integer values ideal for identifying sets of pixels. For a bad pixel mask the good pixels have a value of zero and bad pixels have positive integer values.

The second type is any image format. The image will be internally converted to a bad pixel mask. Note that real image values will be truncated to integers. Again good pixels will have values of zero and bad pixels will have positive values.

The final format is a text file with lines giving the integer coordinates of a single pixel or a rectangular region. A single pixel is specified by a column and line number. A region is specified by a starting column, an ending column, a starting line, and an ending line. Internally this file is turned into a bad pixel mask of the size of the input image with values of zero for non-specified pixels, a value of two for pixels with narrowest interpolation direction along lines, and three for pixels with narrowest interpolation direction along columns.

As noted previously, bad pixels are "fixed" by replacing them with values by linear interpolation to the nearest pixels not identified as bad. Normally interpolation is performed across the narrowest dimension spanning bad pixels with interpolation along image lines if the two dimensions are equal. However specific values in the mask may be used to identify the desired interpolation direction. The value in the mask specifying line interpolation is given by the linterp parameter and the value specifying column interpolation is given by the cinterp parameter. Any values which are do not match one of these parameters results in interpolation along the narrowest dimension. Note that a text file mask always has 1 for pixels with narrow dimension along lines and 2 for pixels with narrow dimension along columns.

The verbose allows printing a line giving the task name, the image name, and the mask name. In addition, if the pixels parameter is set the pixels modified are listed. The list of pixels consists of the column and line of the pixel, the original and replaced pixel values, and the column and line of the one or two pixels used for the interpolation. If the bad pixel region has no pixels at one end, that is there are bad pixels all the way to one edge of the image, then the single pixel used is printed.

Normally the input images and the masks will have the same dimension. However, this task matches bad pixels in the masks with the input images based on physical coordinates. Thus, the mask image may be bigger or smaller than the input image and image sections may be used with either the input images or the bad pixel mask or image mask images. If the input image is the result of extracting a subsection of a bigger image the coordinates of the pixels will be those of the original image and the matching coordinates of the mask will be applied. This has the effect of allowing image sections to be applied to images having a bad pixel mask specified in the image and still having the bad pixel mask be valid.

Mask images may be made in a variety of ways. Any task which produces and modifies image values may be used. Some useful tasks are imexpr, imreplace, imcopy, and mkpattern. If a new image is specified with the explicit ".pl" extension then the pixel mask format is produced. Two other ways to make masks are with the tasks text2mask and ccdmask. The former uses an input text file consisting of rectangular region. This is the old "fixpix" format. The task ccdmask is specialized to make a mask of bad pixels from flat fields or, even better, from the ratio of two flat fields of different exposure levels.

Examples

1. A list of images have bad pixel masks defined in the image header. To replace the bad pixels by interpolation along the narrowest dimension:

cl> fixpix obj* BPM

2. A simple mask of 0s and 1s defines bad columns in spectral data with dispersion along the lines. To interpolate along the lines:

cl> fixpix spec00*h ccdmask linterp=1 v+
FIXPIX: image spec001.imh with mask ccdmask
FIXPIX: image spec002.imh with mask ccdmask
...

3. A text file of bad pixels is used and the modified pixels are printed with:

cl> type mask.dat
1 2 1 1
25 26 25 25
26 27 27 27
49 50 50 50
10 10
20 21 20 20
cl> fixpix myimage mask.dat v+ p+
FIXPIX: image myimage with mask mask.dat
   1    1       1.       1.   1    2
   2    1       1.       1.   2    2
  10   10       1.       1.   9   10  11   10
  20   20       1.       1.  20   19  20   21
  21   20       1.       1.  21   19  21   21
  25   25       1.       1.  25   24  25   26
  26   25       1.       1.  26   26  26   28
  26   27       1.       1.  26   26  26   28
  27   27       1.       1.  27   26  27   28
  49   50       1.       1.  49   49
  50   50       1.       1.  50   49

4. Because a text file input automatically sets the mask values to 2 or 3 you may need to set the linterp and cinterp parameters to force the direction. In particular, to apply FIXPIX to a 1D image, say a spectrum, if you have regions described by ranges in columns the mask interpolation values will be assigned as 3. This means it is trying to interpolation between line 0 and line 2 which is obviously not what is intended. To make this work set linterp to 3:

cl> fixpix myimage mask.dat linterp=3

Revisions

FIXPIX V2.11: This task replaces the old task (now obsolete.ofixpix) and works with the more general pixel mask facilities. It also provides greater flexibility in choosing the interpolation direction.