ccderaser(1) General Commands Manual ccderaser(1)
NAME
ccderaser - erase X-rays, defects and fiducial markers from images
SYNOPSIS
ccderaser [options] input_file [output_file]
DESCRIPTION
This program replaces deviant pixels with interpolated values from sur-
rounding pixels. It is designed to correct defects in electron micro-
scope images from CCD cameras. It can use two algorithms to automati-
cally remove peaks in intensity caused by X-rays. It can also take an
IMOD model file with specifications of regions to be replaced; in this
mode it can be used to erase gold fiducial markers. With a model, the
program can replace a group of adjacent pixels with interpolated val-
ues, or all of the pixels along a line. It can do this on only a spe-
cific image, or on all of the sections in the file. The program can
operate in trial mode, without making an output file, and it can output
a model file with points at the pixels to be replaced.
Automatic X-Ray Removal
The automatic removal of X-rays works by dividing the area of each
image into patches for scanning. The mean and standard deviation (SD)
of the pixels in a patch are computed. The patch is then scanned for
pixels that deviate from the mean by more than a criterion number of
SDs (the scan criterion, a relatively low number to keep from missing
peaks). When such a pixel is found, the program searches neighboring
pixels to find a peak in intensity. It then computes the mean and SD
of pixels in an annulus around the peak and makes sure that the peak
deviates from this local mean by more than a criterion number of SDs
(the peak criterion). Neighboring pixels inside the inner radius of
the annulus are added to the list of pixels to be replaced if they
deviate by a lower criterion (the grow criterion). The patch of pixels
is then replaced by fitting a polynomial to adjacent pixels and inter-
polating from the polynomial. If the peak does not deviate suffi-
ciently from this local mean, but is stronger than the mean of the scan
area by the scan criterion plus 1, then the mean and SD is again com-
puted in a larger annulus. If the peak deviates from this mean by a
number of SDs bigger than another criterion for extra-large peaks, a
patch of pixels is found, but it is replaced only if enough pixels dif-
fer from adjacent ones by large enough amounts (see the -big option
below). The reason for these two stages is that the inner radius for
the first stage must be set safely smaller than the radius of gold
beads to avoid erasing part of the beads, whereas the second stage can
work with larger areas because it has more stringent criteria that
reject gold beads.
After the peaks are found in a scanning patch, the program next finds
the difference between each pixel and the mean of the eight adjacent
pixels. The mean and SD of this difference is computed, then pixels
are sought that deviate from the mean by yet another criterion, the
difference criterion. When such a pixel is found, neighboring pixels
are examined and added to the patch of pixels to replace if their dif-
ference exceeds the grow criterion. If the number of pixels in the
patch does not exceed a specified maximum, replacement proceeds as
above; otherwise the patch is ignored.
Two methods are used because the first method is probably more reliable
for dealing with strong peaks that extend over several pixels, while
the second method is definitely better for finding small X-rays.
After all the patches have been scanned for a section, the program then
searches for single pixels with large interpixel differences at the
edges of the image, over the width set by the -border option. A dif-
ference between a pixel and the mean of whatever adjacent pixels exist
is computed and its deviation from the overall mean interpixel differ-
ence is divided by the maximum SD of interpixel differences over all of
the scans. When this value exceeds the difference criterion and the
interpixel difference is greater than that of its neighbors, the pixel
is replaced with the mean. This procedure is iterated up to 4 times to
catch adjacent extreme pixels.
Tuning the removal of X-rays would primarily involve adjusting two of
the criteria. The peak and difference criteria would be adjusted down
or up to increase or decrease the number of deviant pixels that are
found. The grow criterion could also be adjusted down or up depending
on whether too few or too many pixels are included in a patch that is
replaced, but this step is not usually done in practice. If there are
strong, large artifacts that are not being removed, the big difference
criterion for extra-large peaks should be lowered first, then if neces-
sary, the maximum radius and criterion strength for extra-large peaks
can be adjusted.
Manual Removal of Defects and Markers
For manual removal of defects, prepare an IMOD model file to specify
the points to be replaced. There can be four kinds of objects:
1) Pixel objects, in which there is a point inside of every pixel to
be replaced, and each contour corresponds to a separate patch of
points.
2) Boundary objects, in which each contour encloses a separate patch of
pixels to be replaced.
3) Line objects, in which each contour defines a horizontal or vertical
line of points to be replaced.
4) Circle objects, used to replace pixels in a circle around each
point.
Pixel, boundary, and line objects can be applied either on the section
where they are drawn, or on all sections. For example, one might have
two different pixel objects, one for patches to be replaced on a single
section and one for patches to be replaced on all sections. Note that
if you are working with a montage, "all sections" means "all pieces".
If you are modeling on images displayed as a montage with no overlap
between the pieces, you only need to draw a patch on one of the pieces
to have it erased on all the pieces at each view.
TO set up a pixel object, make the object type be scattered points, and
turn on drawing of circles, but leave the sphere radius at zero. To
specify a patch of points, start a new contour and place a point inside
of EVERY pixel in the patch. Each separate patch should be in a sepa-
rate contour. If the correction is to be made on only a single sec-
tion, all of the points must lie on that section; but the points for a
patch to be corrected on all sections can be on more than one section,
because sometimes they are more discernable on particular sections.
A boundary object should be the default, closed contour type. Draw a
contour around each patch of pixels to be erased. A pixel will be
included in the erasure if its center is inside or on the boundary
line. Contours can be arbitrarily large, and there is tapering of
intensities inside contours that are bigger than 1000 pixels.
A line object should be closed or open contour type. To specify a line
of points, make a contour with two points, at the start and end of the
line to be replaced. Lines must be horizontal or vertical. Each pixel
will be replaced by the average of the two pixels on either side of the
line. Put each line in a different contour. Ajdacent lines will be
detected and erased together.
To set up a circle object, make the object type be scattered points,
set the sphere radius to an appropriate value, and check the option to
display spheres only on their center section. If necessary, you can
give some of the points sizes that are different from the default by
opening a dialog with Edit-Point-Size. A pixel will be replaced if its
center lies inside or on the circle. For small radii, the fact that
the default sphere radius is constrained to an integer may be a prob-
lem; if so, use the -better option to enter a floating point default
radius for Ccderaser to use. For using circle objects to remove gold
markers, -merge should also be used so that adjacent markers are
removed together, and -exclude will help keep dark pixels just outside
the circle from making the filled in region too dark.
OPTIONS
Ccderaser uses the PIP package for input (see the manual page for
pip) and can take input interactively only for specifying manual
replacement with a model file, to maintain compatibility with old com-
mand files. The following options can be specified either as command
line arguments (with the -) or one per line in a command file or param-
eter file (without the -). Options can be abbreviated to unique let-
ters; the currently valid abbreviations for short names are shown in
parentheses.
-input (-in) OR -InputFile File name
Input image file. This may also be entered after all arguments
on the command line.
-output (-outp) OR -OutputFile File name
Output image file. This may also be entered after an input file
name on the command line. If no output file is specified and
the program is not run in trial mode, pixels will be replaced in
the input file. USE REPLACEMENT OPTION WITH CAUTION.
-halffloat (-ha) OR -HalfFloatModeOutput Integer
Enter 2 to set the mode for output files to half-floats (16-bit
floats) unconditionally, or 1 to do so only when the input file
is floating point.
-piece (-pi) OR -PieceListFile File name
File with piece coordinates for a montaged image file. With
this entry, the program will assume that an input model was
drawn on a display of the montage with overlaps adjusted to be
either 0,0 or the amount indicated by the -overlaps option. The
output model file with points found will also be formatted to
fit on such a display.
-overlaps (-ov) OR -OverlapsForModel Two integers
Overlaps in X and Y at which a montaged image is being viewed
both when producing an input model and viewing the output model
of pixels being replaced. The same values should be used as are
provided to the -o option of 3dmod. The default is 0,0, which
represents no overlap between images.
-find (-f) OR -FindPeaks
Find and erase peaks a criterion # of SDs above or below back-
ground. This option must be included for automatic removal of
X-rays.
-peak (-pe) OR -PeakCriterion Floating point
Criterion # of SDs above local mean for erasing peak based on
intensity (the default is 10 SDs).
-diff (-d) OR -DiffCriterion Floating point
Criterion # of SDs above mean pixel-to-pixel difference for
erasing a peak based on differences (the default is 10 SDs)
-grow (-gr) OR -GrowCriterion Floating point
Criterion # of SDs above mean for adding points to peak (the
default is 4 SDs).
-scan (-sc) OR -ScanCriterion Floating point
Criterion # of SDs of mean of scan area for picking peaks in
initial scan (the default is 3 SDs).
-radius (-r) OR -MaximumRadius Floating point
Maximum radius of peak area to erase (the default is 2.1 pix-
els).
-giant (-gi) OR -GiantCriterion Floating point
Criterion # of SDs above local mean for erasing extra-large peak
based on intensity (the default is 12 SDs). Enter 0 to disable
this second stage of peak removal.
-large (-la) OR -ExtraLargeRadius Floating point
Maximum radius of extra-large peak area to erase (the default is
8 pixels).
-big (-bi) OR -BigDiffCriterion Floating point
An extra-large peak will be erased only if the value for the
maximum difference between adjacent pixels, averaged over the
most extreme one-fourth of the pixels in the patch, exceeds this
criterion, evaluated as the number of SDs above the mean abso-
lute difference between adjacent pixels in the scan area. The
default is 19. This high a value is needed to prevent gold era-
sure on low-noise data sets with small gold particles, and a
lower value may be needed to make extra-large peak removal use-
ful.
-maxdiff (-ma) OR -MaxPixelsInDiffPatch Integer
Maximum number of pixels in a patch that exceeds the difference
criterion. If the number of pixels that satisfy the criteria
exceed this limit, none of the pixels in the patch will be
replaced. The default is 2.
-outer (-oute) OR -OuterRadius Floating point
Outer radius of annulus around a peak in which to calculate
local mean and SD (the default is 4.1 pixels). This option and
-width are mutually exclusive.
-width (-w) OR -AnnulusWidth Floating point
Difference between outer and inner radius of the annulus around
a peak in which to calculate local mean and SD (the default is
2.0 pixels). This option and -radius are mutually exclusive.
-xyscan (-x) OR -XYScanSize Integer
Size of regions to compute mean and SD in for initial scans (the
default is 100 pixels).
-edge (-ed) OR -EdgeExclusionWidth Integer
Width of area to exclude on all edges of image in pixels
(default 0).
-iterations (-it) OR -SearchIterations Integer
Number of times to iterate search for peaks. For a given sec-
tion, the iterations will be terminated after an iteration with
no changes. Moreover, the program will keep track of which scan
regions have changes on each iteration and stop redoing regions
that have had no changes. The default is 3.
-points (-po) OR -PointModel File name
Output model file with points replaced in peak search. Each
patch of points will be in a separate contour. Patches are
sorted into separate objects based on how much the peak exceeds
a criterion.
-model (-mo) OR -ModelFile File name
Input model file specifying points or lines to be erased.
-lines (-li) OR -LineObjects List of integer ranges
List of objects that define lines to be replaced. Ranges can be
entered, and / to specify all objects.
-boundary (-bou) OR -BoundaryObjects List of integer ranges
List of objects that contain boundary contours around pixels to
be replaced. Ranges can be entered, and / to specify all
objects. Boundary contours can be arbitrarily large. When a
contour encloses more than 1000 pixels, a different procedure is
use to erase the points, and intensities are tapered down over 8
pixels on the inside of the patch from the value of the nearest
pixel just outside the patch to the mean of the surrounding pix-
els.
-allsec (-a) OR -AllSectionObjects List of integer ranges
List of objects with points to be replaced on all sections.
Ranges can be entered, and / to specify all objects.
-circle (-c) OR -CircleObjects List of integer ranges
List of objects that contain scattered points for replacing pix-
els within a circle around each point. The sphere radius, which
can be an individual value for each point, is used to indicate
the size of circle to replace. Ranges can be entered, and / to
specify all objects.
-better (-be) OR -BetterRadius Multiple floats
For circle objects, this entry specifies a radius to use for
points without an individual point size instead of the object's
default sphere radius. This entry is floating point and can be
used to overcome the limitations of having an integer default
sphere radius. If there are multiple circle objects, enter one
value to apply to all objects or a value for each object.
-expand (-exp) OR -ExpandCircleIterations Integer
Number of iterations of a procedure that expands the area erased
around the points of a circle object. This procedure analyzes
the intensity of pixels adjacent to the patch being erased and
adds pixels in the appropriate tail of the intensity distribu-
tion to the patch. Iterating simply repeats the procedure. If
edges of gold particles are left because points are not very
well centered, try this option instead of just increasing the
radius, as it will tend to erase just the remaining points on
the edge instead of a whole ring of points.
-merge (-me) OR -MergePatches
Merge patches in the model if they touch each other, as long as
the resulting patch is still within the maximum radius. Patches
from objects with points to be replaced on all sections are
ignored. This option should be used if an output model from
automatic peak finding is modified and used as an input model.
-skip (-sk) OR -SkipTurnedOffPoints
Skip points in the model with contour values below threshold if
they are not visible in 3dmod, namely if "Turn off below thresh-
old" is set in the Bead Fixer, or "Turn off Low" is set in the
Values panel of the Model View Object Edit dialog.
-border (-bor) OR -BorderSize Integer
Size of border around points in patch, which contains the points
which will be fit to (the default is 2 pixels)
-order (-or) OR -PolynomialOrder Integer
Order of polynomial to fit to border points. The default is 2,
which includes terms in x, y, x**2, y**2 and x*y. The order can
be between -1 and 3, where 0 will simply replace the pixels with
the mean of the border points instead of fitting to them, and -1
will fill the area with noise matching the standard deviation in
surrounding pixels.
-exclude (-exc) OR -ExcludeAdjacent
Exclude points adjacent to patch points from the fit; in other
words, compute the polynomial fit to points that do not touch
the ones being replaced.
-trial (-t) OR -TrialMode
Analyze for replacement points without writing output image
file.
-verbose (-v) OR -Verbose
Print details on patches being replaced
-PID (-PI) OR -ProcessID
Output process ID
-param (-pa) OR -ParameterFile Parameter file
Read parameter entries as keyword-value pairs from a parameter
file.
-help (-he) OR -usage
Print help output
-StandardInput
Read parameter entries from standard input.
INTERACTIVE INPUT
If the program is started with no command line arguments, it reverts to
interactive input with the following entries:
Input image file
Output image file, or <Return> to place modified sections back into the
input file. USE REPLACEMENT OPTION WITH CAUTION
Model file
A list of objects which specify points or lines to be replaced on all
sections, or / if all objects do so, or Return if none do. Ranges may
be entered.
A list of objects which specify lines to be replaced, or / if all
objects do so, or Return if none do. Ranges may be entered.
Size of the border around the points in a patch, which contains the
points which will be fit to (/ for default of 2 pixels)
Order of polynomial (/ for default of 2, which includes terms in x, y,
x**2, y**2 and x*y)
0 to exclude or 1 to include points adjacent to the points being
replaced in the polynomial fit (/ for default of 1)
HISTORY
Written by David Mastronarde 11/10/98
Automatic X-ray removal and pioneer use of PIP input, 6/9/03
BUGS
Email bug reports to mast at colorado dot edu.
IMOD 5.0.2 ccderaser(1)