imodmesh(1)                 General Commands Manual                imodmesh(1)



NAME
       imodmesh - create a triangle mesh from IMOD contour data.

SYNOPSIS
       imodmesh  [options]  model_files...

DESCRIPTION
       imodmesh is used to create mesh data in an IMOD model file from closed
       or open contours.  Each contour is connected to contours directly above
       and below by an algorithm that minimizes the total area of the trian-
       gles used to connect the contours.  Branches in a volume are handled by
       combining contours in the same plane with additional points.  Contours
       contained totally within other contours are considered to represent
       inside surfaces and will be connected to the outer surface where appro-
       priate.

       Note that "surface" is used here to refer to physical surfaces as com-
       monly understood, and "Surface" (capitalized) refers to contours
       assigned a Surface number in the IMOD model.


   Open and Closed Contours
       Most modeling of structures in a plane is done with closed-contour
       objects, where the brightly lit surface is on the outside of the con-
       tours.  An open-contour object can be used instead to represent an
       extended sheet or a structure that is cut open on one side.  In this
       case, there is no obvious way to define the inside and outside of the
       contours, and the program will assign the outside surface based on the
       direction that contours are drawn.  Specifically, the outside of a sur-
       face formed by a connected set of contours is determined by direction
       of the contour at the lowest Z in that set; higher contours are auto-
       matically inverted if necessary to maintain a consistent outside face.
       If you are facing the surface from the side that you want to be
       brightly lit, the lowest contour should progress (point number should
       increase) from left to right.  This rule will cover most situations,
       but if you are skipping sections in an irregular fashion that requires
       multiple passes to form all connections, then you need to use a consis-
       tent direction for the lowest contour in each set of contours that are
       connected on the first pass.  If a contour is going the wrong way, you
       can correct it in 3dmod with Edit-Contour-Invert.  If the entire sur-
       face is facing the wrong way, you can invert the polarity of the sur-
       face at the level of the whole object; use Edit-Object-Type to define
       the "Front Surface" as being "Inside" rather than "Outside".  You can
       also turn on "Light both sides" in the Model View Object Edit dialog to
       avoid lighting direction issues.

       When meshing open contour objects, imodmesh will attempt to connect all
       contours between sections, regardless of proximity, because it cannot
       use overlapping areas to assess whether contours should be joined.  It
       will pair up the closest contours first, which should lead to correct
       results in most cases.  If incorrect connections occur, contours can be
       assigned to different Surfaces in 3dmod, and the -S option can be used
       to mesh only contours within each Surface.

       If only some of the contours in an object need to be open, such as if
       the object is only partially cut open, then individual contours in a
       closed-contour object can be defined as open with the Edit-Contour-Type
       window in 3dmod.  Imodmesh will properly join these open contours to
       each other and to the closed contours in the same object.  The inside
       of one of these open contours is defined in the same way as if the con-
       tour were closed, and overlap of enclosed area is used to assess
       whether such contours should be connected, so using open contours in a
       closed-contour object may give superior results to using open-contour
       objects.  Two cautions should be noted when using open contours in
       closed-contour objects.  First, when there is a branch involving an
       open contour, the open contour will be joined with one or more closed
       contours on the same section to form a merged open contour that pre-
       serves the open ends; but if such branching requires two open contours
       on the same section to be joined, the surface will not be meshed cor-
       rectly.  The way around this is to use gaps insteads of open contours
       (see "Gaps and Phantom Extensions" below).  Second, if the line con-
       necting the start and end of an open contour would intersect the rest
       of the contour, difficulties will arise if that contour has to be
       joined with another on the same section.  The solution to this problem
       is to use invisible extensions to make the contour encompass the whole
       area with contours in the object (see below).


   Caps
       A cap is used to close off a surface at a place where a contour is not
       connected to another contour on the adjacent section.  Imodmesh will
       create such caps when you use the -c or -C option.  If the unconnected
       contour is relatively round and convex, imodmesh inserts a single point
       above or below the contour, and connects that point to all of the
       points in the contour.  If the contour is more than 1.5 times as long
       as it is wide, or if it is concave by a significant amount, the program
       will analyze the contour to determine the cap, essentially shrinking it
       to a single, potentially branched line.  This procedure, contributed by
       Lennert Ploeger, is new in IMOD 4.12.11 and should greatly reduce the
       need to create your own caps over complex structures.  There are still
       several situations in which you may want to create your own caps by
       hand, such as if only some of the unconnected contours on a section
       should be capped, or if the cap created by imodmesh is inadequate.  To
       create a simple cap by hand, just create a contour with one or two
       points either directly above or directly below the one you want capped.
       To create a complex cap, for example to cap off a U-shaped contour,
       create a contour on the adjacent section with as many points as needed
       to describe the cap, then use Edit-Contour-Loopback to make the contour
       be a complex cap that encloses no area.  The automatic complex cap will
       not work over nested contours (see next section). To place a cap over a
       structure with an inside contour (e.g., a donut-shaped structure), draw
       a single closed contour over the area between the outer and inner con-
       tours that is being closed; this contour should be meshed with both the
       outer an inner contours.


   Inside Contours
       A contour that is completely contained inside another contour of the
       same object is assumed to describe a surface that faces in the opposite
       direction.  Typically, this means that an inner contour is assumed to
       face inward rather than outward.  In fact, contours may be nested to
       any level, with the outermost contour facing outward, the next one in
       facing inward, the next one in facing outward, etc.  Connections will
       be made appropriately, i.e., only between contours that are facing the
       same direction.  In addition, three kinds of transitions can be han-
       dled.  First, if there is a transition from an annulus (one contour
       inside another) to a single, simple contour that encloses all of the
       area covered by the inside contour on the adjacent section, then the
       inside surface is assumed to end and will be capped, if caps are being
       made.  Second, if there is a transition from an annulus to a single
       contour that is shaped like a horseshoe, and does not enclose the area
       over the inside contour, then the horseshoe-shaped contour will be
       divided into an inside and an outside contour for the purpose of con-
       necting to the two contours on the adjacent section.  This transition
       is tricky and may not be handled correctly, especially if more than one
       inside contour is involved in the transition, or if two such transi-
       tions occur in close proximity.  If problems arise, try to avoid having
       more than one transition occur across a given pair of sections, and
       make the points at the tip of the horseshoe (where the contour will be
       divided into two) be close together.  Finally, a branch may emerge from
       the inside of an inside contour just as it may from a normal, outside
       contour.

       Capping over these complex contours will generally be made correctly by
       the new skeletonization procedure.  Horseshoe-shaped contours will cap
       well unless the points at the tips of the horseshoe are too close
       together.  A loopback contour can be added if this capping fails.
       There is no manual way to place a cap right over an annulus, but the
       meshing routine will try to cap this structure by breaking it into a
       horseshoe at its narrowest point and finding the capping contour for
       that modified contour.  There will be a small dent in the cap where
       this break was made.


   Fine-grained Display Information
       Imodmesh will encode some fine-grained display information into the
       mesh, specifically, color, transparency, 3D width changes, and stored
       values.  In addition, 3D width, point size, or symbol size can be used
       to set the diameter of a tubular mesh around open contours.  Contour-
       specific and point-specific properties will always be encoded into the
       mesh.  This means that you cannot see changes in these properties with-
       out remeshing the model.  The treatment of Surface-specific properties
       depends on whether the -S option is entered to connect contours only
       within Surfaces.  If this option is not given, then Surface-specific
       properties are encoded in the mesh as well, and changes in these prop-
       erties will not be seen unless the model is remeshed.  If the option is
       given, then the Surface-specific properties are not included in the
       mesh but are applied during display and can be changed dynamically.

       Connection numbers can be assigned to either points or contours and
       Imodmesh will use these numbers when making the mesh.  Point connection
       numbers can be used to force the mesh to join particular pairs of
       points between two contours.  Most commonly this is needed between con-
       tours on adjacent planes, but it can also be used to specify where to
       connect contours on the same plane that need to be joined to mesh with
       a single contour on an adjacent plane.  See the Fine Grain dialog help
       for more details.  To force a connection between two contours that do
       not overlap, give them both the same connection number.  To prevent a
       connection between contours that do overlap, give them different con-
       nection numbers.  If only one of a pair of contours has a connection
       number it will be ignored when deciding whether to join them.


   Gaps and Phantom Extensions
       A gap in a contour can also be defined to occur after an individual
       point, either with the Fine Grain dialog or with a hot key in 3dmod.  A
       gap at the last point in a closed contour can be used interchangeably
       with defining the contour as open in most situations.  One place where
       the program makes a distinction is when there are other points with
       gaps adjacent to the ends of an open contour.  You can add such points
       in order to make an invisible extension of the contour to enclose other
       parts of the object that need to be meshed with the open component.
       Such added points will be referred to as phantom extensions.  The typi-
       cal case is when you are modeling a sheet (such as a fraction of a cell
       membrane) to which other components such as vesicles attach.  The most
       reliable way to get this case to mesh properly is to add two or three
       points to a contour so that it encloses the entire image on the side of
       the sheet containing the rest of the components in that object, then
       define gaps to make the added segments invisible.

       In order for the program to recognize this situation for a closed-con-
       tour object, the contour must be defined as open, and you must have at
       least two phantom segments adjacent to the opening.  The segments can
       all be at the start, all at the end, or at both the start and the end.

       You may need to add phantom extensions to only one or a few contours.
       If any contours do have phantom extensions, then the extensions will be
       added to all open contours in the object that do NOT have any gaps
       adjacent to the opening, provided that the opening is sufficiently
       large relative to the length of the contour.  For such a contour, the
       program will find the nearest contour in Z that does have extensions
       and simply copy the extensions from that contour.  Thus, you may have
       to draw new extensions at places where the endpoints very differently.
       The automatic attachment of extensions could fail.  If the program
       fails to attach an extension to a contour (because it is very long and
       convoluted relative to the distance between its ends), then you need to
       add extensions to the contour.  If the extensions are added by mistake
       to an open contour that is not part of the sheet, then change the con-
       tour to closed and define a gap at its last point.

       The meshing code will try to match up corresponding gaps between sec-
       tions and connect their endpoints.  Adjacent gaps are combined into one
       for the purpose of connecting endpoints.  Here also open contours are
       treated specially.  Connections are created in three stages:
       1) If both contours are open and there is at least one gap adjacent to
       the opening, the program connects the two openings together.
       2) The pair of remaining gaps with the least distance between their
       midpoints is found and connected provided that this distance is no big-
       ger than the largest gap opening.  This process is repeated until all
       gaps have been considered.
       3) If both contours are open and neither opening has been paired with
       another gap yet, the two openings are connected.
       In each of these cases, the connections are added only if none of the
       points at the ends of the two paired gaps have a connection number
       defined.  Thus, you can prevent an erroneous pairing and connection of
       two gaps by defining a connection number on one of the four endpoints.
       If two gaps fail to be connected by this procedure and artifactual tri-
       angles are added in the gap, then you need to define connection numbers
       at one or both ends of the gap by hand.  See the Fine Grain dialog help
       for more details.


   Tilted Contours
       The program's meshing routines assume that contours are flat in the X/Y
       plane and located at integer steps in Z.  However, it will first ana-
       lyze the contours of an object to find out if they deviate from flat-
       ness.  If they do, it will find an average orientation of the contours
       in each Surface and rotate the contours into the X/Y plane.  It then
       meshes the rotated contours and rotates the mesh back to the original
       orientation to fit the contours.  In order for this to work, two rules
       must be followed.  First, all the contours in a Surface must have very
       nearly the same orientation. (If you do make a small orientation change
       within a Surface, the mesh will be at the average orientation and no
       longer overlay the contours exactly.)   An object can have contours in
       different orientations as long as they are assigned to different Sur-
       faces.  Second, the average spacing between contours must be at least
       one pixel in the direction perpendicular to the plane of the contours.
       This spacing will be maintained when you model in the slicer window
       with the Lock button on and step between slices with the PageUp and
       PageDown keys.


   Stored Parameters
       All of the parameters used to mesh an object are stored in the object.
       Most of these stored values are presented in the Remesh panel of the
       3dmod Model View Object Edit window, where they can be modified and
       used to remesh the object.  By default, the stored parameters will be
       ignored and replaced when you mesh with Imodmesh.  However, with the -u
       option, you can use the stored parameters instead of having to enter
       the options specifying them.  When you do use stored parameters, any
       option entries that you make will override specific stored values.  In
       addition, a set of "-no" options are available for turning off options
       selected in the stored parameters.


OPTIONS
       -u     Use parameters stored in each object to set options for meshing.

       -c     This option will cap off the ends of objects at their extreme
              limits in Z.

       -C     This option will cap off all unconnected contours.

       -D list of Z-values
              This option can be used with the -C option to keep caps from
              being made on certain sections.  The list of Z-values (a comma-
              separated list of ranges) should specify the Z-values at which
              contours are missing for whatever reason.  An unconnected con-
              tour on an adjacent section will not be capped.

       -p value
              Don't make connections unless a given percentage of the contour
              overlaps with contours above and below.  The valid range is 0 to
              100.  The default is 0, which means that any amount of overlap
              will cause a connection to be made.  This option is ignored dur-
              ing the -f option's second pass through the data.

       -s     Normally connections are only made to adjacent sections in the
              image data.  This option allows connections through sections
              containing no data.  Note that with this option alone, the sec-
              tions must contain either no data or whatever is considered to
              be a full amount of data.

       -P value
              Do the given number of passes through the contours, connecting
              contours that are progressively farther apart in Z on each pass.
              In the simplest case, contours up to 2 sections apart will be
              connected with 2 passes, etc.  However, when the -s option is
              also given, the program will connect contours across gaps with
              no data at all on the first pass, then connect contours that are
              even farther apart in Z on the next pass, etc.  To avoid gener-
              ating unintended connections, use both -s and -P and use the
              minimum number of passes needed to make the desired connections.

       -S     This will forbid contours with different Surface numbers from
              being connected.  Contours with the same Surface number will be
              connected between sections, but ones with different Surface num-
              bers will not.  Thus, if you use Surface numbers at all, you
              must make all of the contours that belong in one meshed entity
              have the same Surface number.  Contour connection numbers would
              be an easier way to control a small number of misconnections.

       -I     Ignore time values and connect contours at different times.  By
              default, a model drawn on multiple image files (referred to as
              different times in 3dmod) will be meshed by connecting only con-
              tours with the same time values.  This option can be used to
              override this behavior, in case the contours contain inappropri-
              ate time information.

       -f     Force more connections.  Do a final pass through the contour
              data and make any leftover connections with no requirement for
              contour overlap.  This pass occurs after any multiple passes
              selected with the -P option.  Only contours on sections that
              would ordinarily be connected without the -P option will be con-
              nected.  If this option makes undesirable connections, use con-
              tour connection numbers.

       -t list
              Open contours are by default connected together in a mesh if
              possible.  This option causes a tube-like mesh of diameter given
              by the line width to be created instead, for objects in the list
              (comma-separated list of ranges).  Closed contour objects
              included in the list will be meshed as usual.

       -d diameter
              Override the default diameter used for making tubes with the -t
              option.  The diameter is given in pixels, the default value is
              the 3D line width of the current object.  In addition, -1 or -2
              can be entered to set the diameter from the spherical point size
              or from the symbol size, respectively.  Either of these items
              can vary from point to point.

       -E     Cap the ends of tubes made with the -t option.

       -H     Cap the ends of tubes with hemispheres (domes) instead of flat
              disks.  If you use this option, you do not need the -E option.
              These two options are only way to get tubes capped and are inde-
              pendent of the -c and -C options.

       -T     Do more time consuming calculations by searching exhaustively
              through all possible sets of connections instead of guessing at
              one initial connection.  This may help reduce artifacts in some
              cases.  This flag can cause the computation to take an order of
              magnitude more time.

       -o list of object #'s
              Do operations only on objects in the list of numbers.  The list
              can consist of comma-separated ranges (e.g., 1,7-11,13-17,19).
              Without this option, all objects are scanned for skinning.

       -R value
              Tolerance value for point reduction, given in pixels.  With this
              option, the number of points is reduced by removing ones which
              are less than a certain distance from the remaining line seg-
              ments.  This will remove points selectively from relatively
              straight segments and not from tight corners.  A tolerance value
              of about 0.25 can substantially reduce the number of points with
              little perceptible change in the model; larger values could be
              used to get models that mesh and display faster.

       -i value
              Form a mesh between contours at z-values that are a multiple of
              the given z-increment.

       -z minimum,maximum,[increment]
              Filters which contours will be considered while meshing.  Con-
              tours less then the minimum value and greater then the maximum
              value will be skipped. An optional increment value will cause
              only z-values that are a multiple of the given increment to be
              meshed.

       -x minimum,maximum
              -y minimum,maximum Exclude triangles from the mesh if they are
              outside the given limits in X or in Y.  This is an alternative
              to using clipping planes to chop off surfaces at the edge of the
              data set, although the result will be more ragged than using a
              clipping plane.

       -l     Form a low resolution mesh.  Any new mesh data created by this
              run of the program will be marked as low resolution mesh.
              Existing low resolution mesh will be replaced but existing high
              resolution mesh will be retained.  If tolerance and z-increment
              values are not entered with the -R, -i, or -z options, rela-
              tively large default values are used to generate a coarse mesh.
              The low resolution mesh can be toggled on and off in the 3dmod
              Model View window with the View-Low Res menu entry or the "R"
              hot key.

       -F value
              This option sets the criterion Z difference for deciding whether
              contours are flat enough to mesh without rotating.  If the value
              is nonzero, the program will determine the maximum Z extent of
              all contours in the object.  If the Z difference is greater than
              the criterion, it will mesh each surface separately after find-
              ing the best rotation angles for it.  The default value is 1.5.

       -a     Append the mesh data to objects, replacing only the existing
              mesh in the given range instead of replacing the entire mesh.
              This option would be used to "edit" a mesh that has problems, by
              skinning only a few Z planes with the -T option.

       -e     Erase all mesh data instead of creating new data.  Other options
              besides -l and -o are ignored.

       -n     Recompute the normals in the existing mesh, without computing a
              new mesh from contours.  Other options besides -l and -o are
              ignored.

       -n     Rescale the normals in the existing mesh by the value given in
              the -Z option.

       -Z scale
              A scaling multiplier applied to the z values of normals.  The
              default value is 1.0.

       -noc
       -noC
       -nos
       -noS
       -noI
       -nof
       -noE
       -noT   Turn off the corresponding option when using stored parameters.

       -noD   Remove any restrictions on Z values to cap to when using stored
              parameters.

       -not   Do not mesh open objects as tubes when using stored parameters.

       -nox
       -noy
       -noz   Remove any limits on X, Y, or Z coordinates when using stored
              parameters.

       -B     Make mesh output backward-compatible to IMOD before 3.6.14 if
              possible.  Regardless of this option, if a mesh contains fine-
              grained display changes, then it will be encoded in a new-style
              mesh, which results in a 20% smaller model file.  This option
              can be set automatically for all runs of imodmesh by setting the
              environment variable IMODMESH_OLDMESH.


FILES
       A backup file of the original model is created with the ~ extension.

AUTHORS
       Jim Kremer and David Mastronarde <mast at colorado dot edu>

SEE ALSO
       3dmod, 3dmodv, reducecont

       The algorithm in imodmesh is based on:
       Fuchs, H., Kedem, Z.M., and Uselton, S.P. 1977.  Optimal surface recon-
       struction from planar contours.  Graphics and Image Processing, 20:
       693-702.
       Shantz, M. 1981.  Surface definition for branching, contour-defined
       objects.  Computer Graphics, 15: 242-267.

BUGS
       The algorithm for finding the set of triangles with minimum area is
       currently quadratic; that is, execution time is proportional to the
       square of the number of points in the contours being connected.
       imodmesh can bog down badly if there are too many points; point reduc-
       tion with the -R option will solve this for now.

       Here are some things that will create strange-looking meshes:

       Contours on the same section that overlap.  There is now a function
       that eliminates modest amounts of overlap before meshing.

       Two unconnected branches that overlap in the next section.  Setting a
       threshold for percentage overlap with the -p option may keep them from
       being connected incorrectly.

       Two overlapping horseshoes that open in different places.  The program
       would need to recognize this situation and split both into inside and
       outside contours before trying to join them.



IMOD                                 5.0.2                         imodmesh(1)