Recent Changes In IVE/Priism

Overview

Changes are broken down by the version in which they were incorporated. Recent releases are listed below, for older releases consult OlderChanges.html

1. IVE 4.2.5 (July 18, 2008)
2. IVE 4.2.4 (May 6, 2008)
3. IVE 4.2.3 (November 12, 2007)
4. IVE 4.2.2 (August 21, 2006)
5. IVE 4.2.1 (December 20, 2005)
6. IVE 4.2.0 (September 20, 2004)

Related Priism Topics

Priism | Known problems | Reporting problems | Priism download site

IVE 4.2.5

Features added in 4.2.5 are:

• MatToImg now handles arrays stored as signed integers (8, 16, or 32-bit), unsigned integers (8, 16, or 32-bit), and logicals. The pixel representation for the output is closest match from the available pixel formats used by Priism. ImgToMat now allows selection of how values are stored in the MAT file.
• Added a command-line mode of operation to the Region MMM application.
• (UCSF version only) Added weighting options (equivalent to those in ewbp) to the backprojection algorithm in gpurecon.

Problems fixed in 4.2.5 are:

• Priism 4.2.4 when run on x86_64 Linux systems had some components fail with a symbol lookup error for __intel_new_proc_init.
• The "Print Image" function for image windows incorrectly arranged multiple images on the page when the multiple images in the window were displayed with the lowest section number in the upper left corner (i.e. "text" mode).
• To better handle the case of submitting EM alignment and reconstruction jobs to a remote system where the user's home directory is different than on the machine submitting the job, the front-ends just reference the HOME environment variable in the default value for the temporary directory rather than substituting the current value for the HOME environment variable.
• appl_prm did not set the pixel spacing correctly in the aligned image file when reading from a lower resolution of the input tit series.
• Corrected tapir and reproj: padding added to projection lines would, in some cases, introduce a half pixel misalignment.
• For tapir, report a cross correlation of zero rather than one when there is no data or one of the two compared images has a constant intensity.
• (UCSF version only) gpurecon now prints out the R-factor results from the servers to the log file. Changed gpurecon to avoid hanging when one or more servers have zero lines to reconstruct. Corrected gpurecon so the last column of data from the projections does contribute to the reconstruction. gpurecon permutes the input projections so block-iterative reconstructions do not draw blocks from contiguous ranges of tilt angles.

Other changes in 4.2.5 are:

• Now include libquicktime files from libquicktime 1.0.2 rather than libquicktime 1.0.1.

Return to overview

IVE 4.2.4

Features added in 4.2.4 are:

• Allow modification of the threshold the cross-correlation alignment uses when trying to reduce the influence of a central peak. Added option to write out cross-correlation peak images from the iterative alignment applications.
• Change the event loop routines in the ive library so thread-safety is a decision that can be made at runtime rather than at compile time. Made the event handling routines more flexible by adding the option for work procedures and internally generated events.
• Added IVEGetFreeSpace() and IVEThreadCountProcessors() to the ive library.
• (UCSF version only) Added multiple threads of execution to the GPU reconstruction client so it could overlap communications with the server with disk I/O and computation. Modified the client/server protocol and the client software so a convergence factor can be specified for the SART algorithm and multiple subsets can be specified for the EM algorithm.
• (UCSF version only) Added a version of the Refine Alignment application which makes use of servers with graphics cards to accelerate the reconstructions and projections.

Problems fixed in 4.2.4 are:

• Appending to an existing data set with multiple resolutions with a ProcFunc-based application (CopyRegion, Image Arithmetic, ...) left corrupted data so discard the lower resolutions when appending.
• Made several corrections in the tilt axis determination. For one, print out plot results with more decimal places and allow more space for peak position printed out in plot titles. Also remove pairwise plot result if it already exists to avoid appending to old results and do not generate the pairwise plot results if the file name is "none".
• Corrected the mass-normalization application so multiple runs which fit the tilt offset and use the prior run's parameters as input do not cause the tilt angles to increase with each run. Also supressed output of the baseline curve in the mass-normalization diagnostic plot when the baseline is forced to be zero.
• Using the filtering option in the Refine Alignment application would cause an early exit because of improperly quoted arguments to the filter.
• The tolerances for the fit in the bead alignment scaled inappropriately with the number of marked positions.
• Corrected regression in interactive bead alignment: predicted position of bead not shown in image window for projections where the bead position was not marked.
• Corrected the interactive bead alignment so it is not limited to displaying results for fifty beads.
• Fixed the interactive bead alignment so it would run even if it could not write to a terminal.
• In the bead alignment point editor, now load graphics for all projections rather than loading a section at a time. The change is intended to make scrolling through the projections smoother.
• ccdcor, decon, ffilter and rotfast will now increment the starting time index in the header by the offset for the first time point selected for processing. applycal now increments the starting time index in the header rather than replacing it with the value of the first time point selected for processing.
• In cases where a Gaussian fit to a point did not use three dimensions, Find Points incorrectly set the position for the dimensions that were not included in the fit.
• Find Points would hang in some cases when determining the positions with a Gaussian fit.
• The Slider application now responds to other interactions with the slider besides dragging it with the mouse. Also corrected Slider's behavior when run with a command-line argument specifying a single image window to manipulate.
• On Linux, Image Capture would flip the left and right sides of images when saving as grayscale to a QuickTime movie.
• Changed the handling of scroll wheel (internally button 4 and 5 events) in the image window to consolidate consecutive queued events.
• Increased default value for IVE_WORKING_UNIT in Priism_setup and Priism_setup.sh to tolerate 4096 x 4096 complex data.
• On Mac OS X systems where hostinfo reports the amount of RAM in gigabytes, Priism_setup and Priism_setup.sh would incorrectly set the IVE_WORKING_SET environment variable.
• Added a guard to Priism_setup so it would not clear the path if sourced from a sh or bash shell.
• A dialog size change triggered by a change in the dialog's contents caused an application crash on at least one system (Mac OS X 10.3 with Apple's X11 1.0).
• Changed the implementation of IMRtDel() so it would turn a pixel spacing of zero for a dimension if the mapping of that dimension to rows, columns, or sections was out of bounds. The prior implementation had undefined behavior (including segmentation faults) when the mapping was out of bounds.
• In internal help browser, changed algorithm for locating the topic name to match what web browsers do. Fixes issues accessing some help items from 2d_plot.

Other changes in 4.2.4 are:

• Now use Apple's version of gcc 3.1 when building executables and libraries for PowerPC systems running Mac OS X. Those executables may not be compatible with OS X 10.1.

Return to overview

IVE 4.2.3

Features added in 4.2.3 are:

• Mac OS X and Linux versions now come with LessTif to simplify installation.
• In the EM alignment and reconstruction applications, added the capability to work with a lower resolution version of the projections stored in the same file.
• For the IP framework added an alternate 3D processing mode that takes blocks in x, y, and time rather than x, y, and z. Filter3D, LCE3D, and the 3D Fourier transform applications (FFilter3D, FPwrSpec3D, FTransform3D) make use of this new feature.
• Added 4D versions of the Fourier transform applications: FFilter4D, FPwrSpec4D, and FTransform4D.
• Added a port of IVE 3.3's object library and the Clouds application. Enabled the ability of Pick Points and EditPolygon to generate object library files.
• Added Import/Export->Mrc2Image2000 for converting from Priism's variant of the MRC format to the Image 2000 variant.
• Added DataViews->DataInfo->Edit Header (any MRC) which is similar to EditHeader but able to work with several variants of the MRC format.
• In the tilt axis determination for EM alignment, changed the format of the output graph to be compatible with 2D Plot. Added curves averaging across just the negative and positive tilt angle pairs to the output graph. Added the option to generate a separate file with the cross-correlation as a function of assumed tilt axis orientation for each tilt angle pair.
• In the EM alignment and reconstruction graphical user interface, added an option to display the cross-correlation images and graphs from the tilt axis determination.
• In mrc2ccp4, added an option to set the skew transformation from the rotation angles and origin in the Priism header.
• The converters to Matlab .mat files are now functional on Intel-based Macs.
• Added an option to Straight to allow the user to disable smooth transitions at joints in the model since the process of smoothing breaks a linear relationship between column number in the output images and distance along the model.
• Added a command-line interface for tiff2mrc and extended the file name formats for numbered series of files that tiff2mrc accepts.
• Added two options to the watershed segmentation algorithm: one to check the neighbors in order of intensity rather relative spatial position and another to apply tie breakers to the elements with the same intensity when all the elements are sorted.
• Now use libquicktime 1.0.1 rather than libquicktime 0.9.8 to generate QuickTime movies on Linux.
• (UCSF version only) Added front-ends (the GPURECON option in EMTAR and EMTIARGPU) and client programs (gpurecon and gpureproj for interacting with servers that use graphics cards to accelerate EM reconstruction calculations.
• (UCSF version only) Added a second iterative alignment method for EM reconstruction, Refine Alignment. Each iteration of this method computes a series of weighted backprojections, each of which omits some of the input projections. The projections of a reconstruction are then used to compute the alignment parameters for the projections not used to compute the reconstruction.

Problems fixed in 4.2.3 are:

• The Priism C shell setup script could break when a local variable masked an environment variable.
• Compressed archives of the Priism binaries now specify a top-level directory.
• Corrected the iomenu graphical user interface (used by ccdcor, bead alignment, EM alignment and reconstruction, and others) so that selecting a shorter primary input filename does not leave trailing characters in the script and log file names.
• In 2D Plot, added options to better control the display of tick labels.
• Corrected an off-by-one error in the anisotropic diffusion filter's normalize option.
• In Filter3D, corrected the handling of the statistics in the header when appending to an existing file or window.
• Corrected Find Points so that adjusting the maximum number of points to find after finding points does not leave point markers displayed in an image window.
• Modified FFilter2D, FTransform2D, and FPwrSpec2D to avoid disconcerting scale changes when appending data in z or time to an existing window.
• In FFilter3D, FPwrSpec3D, and FTransform3D, corrected padding with a ramp along the z direction in frequency space.
• Corrected FFilter3D so the output was correct when the z output size is not the same as the z input size.
• Corrected the application of window functions to complex-valued data in FPwrSpec3D.
• Corrected FTransform3D so that it would not ignore a shift that was solely along the z axis. Corrected a problem when applying a shift during an inverse (unpacked to real) transform. Also corrected the handling of the z padding value: previous versions would use the y padding value instead.
• In EM alignment and reconstruction, changed the selection of the xz region of interest so that picking outside the previous region of interest gives a box that can be instantly adjusted with the mouse. Also changed the handling of the default parameter file for xy region selection.
• The documentation for the EM Alignment File Manipulator incorrectly stated the sign convention for shifts.
• Corrected appl_prm so the combination of not applying the alignment parameters and applying a shift or size change to the output works.
• Modified the interactive version of the final stage of bead alignment so errors are correctly adjusted when moving a point near the edge whose fit position has been coerced to fall in the image.
• Removed buffer overflow in the Bead Chaser step of bead alignment when the specified radii of the beads is less than four pixels.
• Changed ewbp and tapir to flag the output reconstructions as normal images; this allows viewing of the reconstruction in Chimera without first changing the image type field in the header.
• Changed tapir to report the agreement between the input projections and projections from the reconstruction using the normal crystallographic formula for the R-factor. What was reported as the "R-factor" in previous versions is also generated to make it easier to compare with old results.
• In the tilt axis determination step for EM processing, corrected the geometric mean calculation across the tilt angle pairs.
• In the tilt axis determination step for EM processing, now exit with an error if no pair of projections matches the criteria specified by the input parameters.
• Corrected an off-by-one error that could cause crashes when generating QuickTime movies on Mac OS X from Image Capture or 3D Model.
• Corrected Line Profile's calculation of profiles in time which averaged over multiple z sections and more than one point in x and y.
• If requested to swap bytes, mrc2ccp4 sets the machine stamp in the file appropriately.
• Rotate3D now sets the starting section index in the header to zero.
• Changed Volume Builder so it can be run without a full Priism session running.
• The watershed segmentation's internal segmentation routine did not adjust the number of regions segmented so far when an early exit was forced due to insufficient storage for the attributes of regions.
• Corrected off-by-one error that affected handling of trailing blanks when WMAddCharField or WMAddScrolledText are called from Fortran.
• Modified the WM library so WMProcEvent will handle Xt timer, input, or signal events when the wait argument is zero.

Other changes in 4.2.3 are:

• Removed versions of archive libraries (libimcompat.a, libiomenu.a, libifft.a, and others) that were compatible with version 7 of the Intel compilers for Linux.

Return to overview

IVE 4.2.2

Features added in 4.2.2 are:

• Provide native binaries for x86_64 Linux systems and for Intel-based Mac OS X systems.
• In the Processing->Enhancement menu, added A. N. Diffusion, an implementation of an anisotropic nonlinear diffusion filter.
• In the Import/Export menu, added Mrc2Spider, an application to convert Priism's native image format to Spider's native image format, and Mrc2CCP4, an application to convert Priism's native image format to CCP4's native image format.
• Introduced a configuration file to configure queues and default parallel execution settings.
• For cluster processing with the parallel EM scripts, now allow input and output files to reside in places not visible to all systems in the cluster. Because of the change, the command-line options for cluster processing are not backward-compatible with Priism 4.2.0 or 4.2.1.
• Improved the performance of the weighted backprojection and iterative reconstruction applications.
• In the application to align EM data from fiducial markers, the option to fix the z position of the fiducial markers now does just that without additional side effects. There are now additional options so you explicitly turn on the additional behavior that in past versions happened automatically when fixing the z positions of the markers.
• Modified the IP framework and most other applications so that when generating an image window with a single z section but multiple time points, the default step direction is along time.
• Applications based on the IP framework now have the option to allow the user to select which resolution to process.
• Use libquicktime rather than Quicktime 4 Linux to generate Quicktime files on Linux systems.
• Changed the default help browser on IRIX and Linux to a wrapper script that detects whether Firefox, Mozilla, or Netscape is available.

Problems fixed in 4.2.2 are:

• Fixed 2D Plot to prevent "Line X of plot file was too long" messages.
• Changed 3D Model so changes made in the "WT Display Options" dialog are immediately reflected in image windows. Also changed 3D Model so image windows are updated when the last object is deleted or the point list for the current object is deleted.
• Corrected the interactive final step of bead alignment to prevent a segmentation fault at exit when asked to generate a file with the aligned projections.
• The final step of the bead alignment now accepts beads which only appear on two projections (previous versions erroneously required three).
• Changed CutMask so masks of complex-valued data use 1+0i as the value for regions in the mask rather than 1+1i.
• Corrected the statistics CutMask saves to the output header.
• Removed memory leak in CutMask when the input has an extended header.
• Changed the behavior of the expansion step of the simplex algorithm used for EM alignments: the new variant appears to give slightly more reliable results than the old. Also, added one restart of the minimization.
• Modified the tapir front-end to accept a value of zero iterations (i.e. simply perform a backprojection).
• Because the deconvolution expects transfer functions where the pixel spacing is in spatial units, added an option to FTransform2D and FTransform3D to keep the pixel spacing in spatial units.
• For centered output from FTransform2D, FTransform3D, FPwrSpec2D, and FPwrSpec3D, shuffle the arrays in memory rather than by multiple I/O calls in order to improve performance when writing to NFS disks.
• Corrected FFilter3D to prevent a segmentation fault when generating a filter profile.
• Corrected the lower bound on the time range in Edit Polygon's clear partial dialog.
• Modified Find Points and Region Growing to implement checks on whether opening outputs for images failed.
• Fixed instances where Find Points and Volume Builder would leave object number labels in an image window.
• Improved the initial placement of subdialogs in Find Points, Straight and Volume Builder.
• Modified Find Points and Region Growing so that clearing the name of a region-growing output file disables that output until a name is provided.
• Corrected the bounds imposed on values in FindPoints z and time fields.
• Modified Find Points so the initial z and time steps are one rather than zero.
• Made the automated propagation of polygons in Edit Polygon work in cases where the destination range has only one time point or does not contain the seed section.
• Corrected problems with Pick Points' handling of copying points between wavelengths or time points or using something other than the first wavelength and time point when copying points in the same wavelength and time point.
• Removed a buffer overflow in Region Growing that could occur when asking the user whether or not to overwrite a file.
• Modified the filter option in Scale to restore behavior broken in IVE 3.3.
• Modified Scale to update the histogram display in response to selecting the 3D histogram or filter options.
• Modified Straight to prevent crashes if "show model lines" or "show current point" is selected and the model was read from a file rather than accessed from main memory.
• Modified the cookie mode in Straight to reduce missing slivers when cutting out a large radius about an object's axis.
• Modified Straight to not pass along extended header information when using the straight or tunnel methods.
• In Volume Builder, corrected the handling of removing the last polygon from an object, removing sibling polygons from an object, and adding polygons to an object with no polygons.
• Tightened bounds checks on values entered in Volume Builder's current object field.
• Modified Volume Builder to display sibling polygons in the correction dialog.
• Changed Volume Builder so that, when renaming the current object, the current object will refer to the renamed object.
• Corrected Volume Builder's highlighting of objects when the user selects a new current object.
• Changed image windows to honor the synchronization of z and time when stepping in wavelength.
• In the IP framework, corrected a consistency check on the z and time dimensions that was too loose.
• Modified the IP framework so the NEW_REGION callback is invoked if a change to whether or not an input derives its bounds from the first input causes a change to the region processed. Also modified the IP framework so the WAVE_FUNC callback is invoked if a change to whether or not an input derives its bounds from the first input causes a change to the wavelengths processed.
• Modified IPGetRegionInfoExt to return correct results when processing is not in progress and the caller requests information about an input which uses the same region as the first input. Made it clear in the documentation when the values returned by IPGetRegionInfoExt are consistent with the dimensions in the input's header.
• Modified the IP framework to prevent a mix of per-section and per-wavelength scaling when the first sections written in a wavelength have a uniform non-zero value while the remaining sections do not.
• Protected against I/O errors when generating scripts from IOMENU-based interfaces.

Return to overview

IVE 4.2.1

Features added in 4.2.1 are:

• Several applications were modified to avoid use of the NAG Fortran 77 library. The applications and the computations affected are:
  • Final stage of bead alignment: fit of alignment parameters and bead positions
  • CCDCor: polynomial fits
  • Decon: computation of focal point changes in depth-dependent deconvolution
  • FFilter2D, FFilter3D, FPwrSpec2D, FPwrSpec3D, FTransform2D, and FTransform3D
  • FindPoints: curve fit to histogram for thresholds and fitting Gaussian profiles for peak positions
  • pffit: fitting Zernike polynomials to pupil function
  • simplex: simplex minimization
  • SurfHarmMod: fit of spherical harmonics to model points
• In the iterative alignment and reconstruction:
  • Built in the preparation step of computing the mass-normalized tilt series. Removed the ability to specify the image formation model and pcbase used when invoking appl_prm in the iterations.
  • Changed the alignment portion to have an option for different alignment methods like the non-iterative alignment. By default, the alignment uses the Talign method and only refines the translations.
  • Added an option (on by default) to terminate iterations when the reconstruction R-factor increases between subsequent alignment+reconstruction iterations.
  • Added an option to carry over the reconstruction between iterations rather than restarting tapir without an initial guess at each iteration.
  • Added the option to specify an initial guess for the reconstruction.
  • Parallelized the alignment portion.
  • Made it easier to use the parallel execution mode on a single machine with multiple processors.
  • Now create the working directory if it does not already exist.
• EM applications no longer generate the .bprm format and always use the .bprmMn format instead. The applications can still read .bprm format input files.
• Changed the front-end to massnorm so the default is to fit the background rather than the mean and to use the entire image for statistics. Made fitting of the baseline intensity level optional; the default now is to not fit the baseline value. Reintroduced option to attempt to identify and correct for abrupt changes in the intensity as a function of tilt angle.
• Changed the defaults for the cross-correlation alignment to be the talign method, .001 for the second peak threshold, and zero for the isotropic magnification characteristic length. Made some adjustments to how the cross-correlation peak search deals with a peak at the origin.
• Modified the scaled linear model in appl_prm to compensate for variations in exposure time. Modified the EM rough alignment to extract the exposure times from the extended header and use them to initialize the scale parameters in the alignment parameter file.
• In the weighted backprojection, modified the handling of the mean (in the new version calculate the mean line by line rather than for a 2D slice through the projections), padding (now use a zero pad in all cases), and the Hamming filtration boundary condition (now wrap around rather than repeating the last element).
• Extended the alignment file manipulator to be able to compensate for image center shifts.
• Modified components in Priism to generally enforce one of two conventions for the sampling frequency entries in the MRC header: either each of the entries is equal to the corresponding image dimension in pixels or all the entries are set equal to one. The EM reconstruction applications always force their output to follow the first convention. To agree with this convention, the Fourier transform and power spectrum modules no longer use the sampling frequencies to store the input size prior to padding and the Fourier transform.
• Changed Resample2D so that, by default, it does not preserve the cell dimensions.
• Changed the Fourier transform and power spectrum modules so the output pixel spacings are each set equal to one divided by the product of the input region size in pixels and the input pixel spacing. In other words, if the input pixel spacings have units of microns, the output pixel spacings have units of cycles per micron. For a forward transform, the Fourier transform and power spectrum modules set the origin coordinates to the frequencies (in the same units as the output pixel spacing) for the pixel in the bottom lower left. For a reverse transform, the Fourier transform modules set the origin coordinates to zero.
• Modified the BlendColors user interface. Selections from the color wheel modify the wavelength selected within that dialog rather than relying on the wavelength selection controls in the main dialog. The lowest priority wavelength control has been integrated into the main dialog. The scaling fields now use the same number of parameters as other applications.
• Added a module for determining the tilt axis orientation to the EM alignment and reconstruction suite.
• Added a parallel version of the weighted backprojection application. Made the parallel versions of the weighted backprojection and tapir accessible from the alignment and reconstruction front end.
• Added an application, IncreaseRes, to function as an inverse for ReduceRes without the full generality and memory overhead of Rotate3D.
• Added an application, insert_tilts, for adding tilt angle information to data sets that do not have an extended header.
• Added a bilateral filter option to Filter2D and Filter3D.
• Added command line interfaces for 2DObjFinder, AppendRes, BlendColors, FPPolygons, ImageCapture, mrc2tiff, mrc2txt, and ViewFile.
• Added option to the bead alignment to hold the tilt axis orientation fixed.
• Added option to Flip to control the amount of working space used.
• Added operations to Image Arithmetic to ease computation of the complex conjugate and multiplication by the complex conjugate.
• Added options in the Priism Color menu to change the image color table. As a side effect, changed the default setting for the image background color to be the first color in the palette of colors used when drawing on the images.
• To reduce problems with unintentionally overwriting a previous log or script file, added a serial number to log and script names generated by IOMENU-based interfaces.
• Made Matlab utilities compatible with Matlab 7 on Mac OS X and x86 Linux. Removed necessity of patching the IRIX version for compatibility with Matlab 6.5.
• Changed the conventions for file selection dialogs so that, with the exception of selecting configuration files, the starting directory is the last directory visited.
• On x86 Linux systems, removed runtime dependence on libglut.so.
• Made Safari the default help browser on Mac OS X.

Problems fixed in 4.2.1 are:

• The Priism_setup and Priism_setup.sh scripts failed to recognize 64-bit x86 systems.
• Modified Priism_setup and Priism_setup.sh to use DYLD_FALLBACK_LIBRARY_PATH rather than DYLD_LIBRARY_PATH: avoids a problem encountered on Mac OS X 10.4 where Priism would refuse to start when /sw/lib was in DYLD_LIBRARY_PATH and the Fink version of Gimp was installed.
• The iterative alignment and reconstruction did not allow specification of the positivity-related options for tapir: it would always coerce the input values to be non-negative but did not apply a positivity constraint to the reconstruction.
• The alignment step in the iterative alignment and reconstruction generated broken .bprmMn files: the column titles were printed on two lines rather than one.
• The iterative alignment and reconstruction ignored the input parameters specifying a translation of the reconstructed volume.
• The iterative alignment and reconstruction interface did not respond to close requests from the window manager.
• The extraction of EM alignment parameters from the extended header had logic errors in how it handled the case where the data set was collected in two chunks and had two different reference points for the shifts.
• Corrected an error in the alignment and reconstruction front end: the r* Hamming option for the weighted backprojection was written out without a leading dash.
• Changed appl_prm and final step of the bead alignment to carry over the start indices (with an adjustment to the z index for the first projection used) from the input data when writing aligned tilt series.
• Corrected errors introduced in the 4.2.0 version of massnorm: fixed to pass along user-specified tilt offsets to the output parameter file, removed use of an uninitialized variable in the read from an input .bprm file, corrected intensity value axis and histogram data in the histogram output, removed access past end of an array that caused the program to crash on Alpha Linux systems.
• Changed the histogram calculations in massnorm so they do not assume the minimum and maximum in the input data's header are correct.
• Changed the weighted back projection and tapir modules to set the start indices and origin to zero in the reconstruction's header.
• Corrected tapir so it did not ignore the smoothing option on Linux.
• Corrected the entries in the "Npos" column of the summary table in tapir's log file. For a particular iteration, the entry in the "Npos" column is the fraction of the reconstructed volume that violated the positivity constraint for that iteration. Corrected the summary statistics for the cross-correlation results: these now reflect the cross-correlation across all sections and not just the last section. When run in parallel, generate cumulative statistics across all chunks.
• Modified tapir to protect against division by zero in the R-factor calculations and in the comparison of the reprojections with the input data when a y line only contains pixels flagged as invalid data.
• Improved the accuracy of the mean value tapir recorded in the reconstruction's header.
• Removed hardwired upper limits on the number of sections from all the EM alignment and reconstruction applications with the exception of those for bead alignment.
• Corrected ccdcor and ffilter to carry over the start indices from the header of the input and to modify the start indices and origin in the header to reflect the region selected for processing.
• In AppendRes, the computation of the imaginary component in lower resolution was incorrect for complex data. Also, the progress indicator was misleading for z downsampling factors greater than one.
• Fixed logic in BlendColors for tracking when dynamically scaled output is open. Corrected a buffer overflow when dynamically rescaling output. Changed to avoid garbling the contents of the graph minimum and maximum field.
• The deconvolution user interface did not respond to close requests from the window manager.
• Protected against crash when the user selected the automatic mode in EditPolygon.
• Corrected the test for whether or not a point is inside a polygon. The change affects EditPolygon and FindPoints.
• Expanded the region outline drawn by Examine Data to include the half pixel wide border that is included in the Examine Data display.
• Protected against a crash in Find Points if no file is open and the user selects "curve fit" from the Method menu.
• Previous versions of Line Profile used a region two times wider than indicated when performing the averaging for diagonal profiles. Removed a one pixel shift in many diagonal profiles. Removed binary characters from output files for diagonal profiles. Modified so image graphics are aligned to the pixel centers. Modified so that when working with multiple windows, plots are saved in order of increasing window number.
• The previous version of RadProj did not correctly handle the radius bounds for data sets whose pixel spacings were not all equal to one. That version of RadProj would also crash when projecting the sum, minimum intensity, or maximum intensity to an image in cylindrical or spherical coordinates.
• Previous versions of ReduceRes crashed when given complex-valued data.
• Fixed simplex to avoid crash when an output image file is requested and there is no second input image file. Removed binary characters in simplex's log output.
• Corrected strait's allocation and freeing of storage buffers when the user selects a new input image window or file.
• Changed Volume Viewer's allocation of buffers so large movies do not cause "IVEShmCalloc fail: Not enough SHM" messages.
• Fixed Volume Viewer so it does not crash when rescaling the input after a preloaded region with a nonzero z starting value has been selected.
• Added check in Volume Viewer to gracefully handle the case of complex-valued data.
• Modified FPPolygons and 2DObjFinder so the default file browser path for polyon files does not interfere with the default file browser path for image files. Modified BlendColors so the default file browser path for color files does not interfere with the default file browser path for images.
• Modified the scripts written by IOMENU-based applications to avoid undue loss of precision for floating-point parameters.
• Corrected problems with the chain option help lookup in IOMENU-based applications.
• Tightened check for a preexisting shared memory file for which the corresponding Monitor 0 process does not exist.
• A misplaced return statement caused image windows to read back scaled image data that had been buffered and then discard that data to recalculate the scaled image.
• Changed the handling of image window scaling so that the accurate scaling mode is always used when the scaling range would be represented by a floating-point infinite or not-a-number value. This prevent image window crashes on the Mac when a scaling range of -inf to 0.5 is used.
• Changed so that image windows report a synthetic window entry event if an application requests window entry events and the pointer is already over the window.
• Fixed some instances where the image window failed to clear the background appropriately.
• Fixed image windows so the indicator for jumping to a section is shown when the visual is not doublebuffered.
• Corrected inconsistencies in the handling of the image background color when it is the same color as the first color in the image color table.
• Protected against crashes in the IW library when unable to connect to the X server.
• Protected against X server hangs when a modal dialog is posted at the same time a pulldown menu is activated.
• Modified get_realswitch to tolerate scientific notation with an "e" or "E" as the prefix for the exponent.
• get_realswitch imposed a limit of 50 characters on the length of the values to be parsed. Increased the limit to 256 characters.

Return to overview

IVE 4.2.0

Features added in 4.2.0 are:

• Restructured the EM reconstruction front-end to include options for getting the alignment parameters from the extended header or to compute them based on cross-correlation results. The mass-normalization step in the back-end no longer attempts to fit jumps and has new options to limit the region of analysis.
• Added another EM reconstruction front-end for iterative refinement of the alignment parameters and reconstruction.
• In Volume Viewer's movie dialog, added an option to perform the incremental rotations relative to the unrotated coordinate system. Changed the rgba/opacity method to use the procedure for scaling the output as the progressive method.
• Added an inverse filtering option to the 2D and 3D Fourier filter applications. Also added a (1-Gaussian)*Gaussian filter like ffilter has.
• Added calculation of the sample standard deviation to the Region Min/Max/Mean application.
• Added the headit application from IVE 3.3 for command-line editing of image file header information.
• Added RadProj, an application for computing radial averages or other projections from Cartesian to polar, cylindrical, or spherical coordinates.
• Added versions of libimcompat, libiomenu, and libifft which are compatible with the IBM XL Fortran 8.1 compiler for Mac OS X.

Problems fixed in 4.2.0 are:

• Corrected a resource leak (one SYSV semaphore per session on Linux) introduced in 4.1.6.
• Added workaround for incorrect drawing of rectangles in image windows on some Imacs (reported for a system using Apple's X11 and an Nvidia FX 5200).
• Fixed loading of multiresolution data with ViewFile to avoid memory corruption if the data does not have the host's native byte ordering.
• Removed cause of UtRmImgMem messages when switching from stereo to non-stereo mode in Volume Viewer.
• Fixed the EM reconstruction front-end so the "Change file names" button is reliable on Mac OS X.
• Stretching was applied incorrectly for the display and modification of the xy region of interest parallelogram.
• Fixed an off-by-one error in the EM reconstruction front-end that caused an erroneous message to be displayed when picking the xy region of interest using the last projection.
• Fixed a logic problem in appl_prm that caused it to use uninitialized values with the scaled linear image formation model.
• Fixed appl_prm so it gets the correct tilt angles when using the linear image formation model and parameters from a .bprmMn file.
• Modified tapir to accommodate y step sizes other than one. Fixed pixel size set in appl_prm's output when the y step size is not one.
• Modified tapir so that, if there is no -istrfile option, it assumes there is no initial guess.
• Removed limit of 50 markers per section in the bead_chaser and bead_align steps of bead alignment.
• Fixed a problem in the bead chaser step of bead alignment that could cause the application to crash.
• Fixed the bead alignment step so the option to output an aligned tilt series is functional.
• Fixed a problem introduced in 4.1.8 that caused the deconvolution front-end to crash when an input file was not specified or when unable to generate the script file.
• The 4.1.9 change to Volume Viewer for setting the output size did not correctly factor in the zoom factor.
• Volume Viewer was incompatible with the 3.2 VLI library for the VP1000 board.
• To facilitate modeling or erasing with Volume Viewer's slicer mode, the slicer mode now uses the central slice rather than end slice from the rendered volume.
• Fixed problem that caused uniformly bright images in Volume Viewer's slicer mode when the volume rendered does not intersect the source volume.
• Fixed Volume Viewer to update the input scaling dialog if a new input file is selected while that dialog is open.
• Modified the 3D cursor to not reset the z mode on window entry to reduce the opportunity for the z mode to get out of sync with Volume Viewer. Changed the z mode indicators in Volume Viewer and the 3D cursor to buttons so it is possible to manually bring the two in sync.
• Modified the 3D cursor to avoid missing window exit events.
• Fixed the 2D and 3D Fourier transform applications so it is not necessary to specify the output format when doing a forward transform of real data. Corrected an error in those applications when limiting what output pixel formats are possible.
• Corrected the minimum, maximum, and mean values written to the header for masked data and region label data generated by FindPoints or Region Growing.
• The command to print an image window now includes the scale bar label in the printed output. However, because the font in the printed output does not match the font used on the screen, the output falls short of "what you see is what you get".
• For complex data written by MatToImg, changed the statistics saved to the header to only reflect the real components (as CopyRegion does).
• Modified the ffilter command-line application so it is possible to inverse filter with an OTF.
• Corrected the ffilter front-end so the conditions for enabling and disabling the bandpass and smoothing/enhancement filters are the same as in the underlying command-line application.
• Fixed the simplex application to not crash at exit if one image input is used.
• Corrected the handling of "%%" in the file name and label templates used by ImageCapture, mrc2tiff, and mrc2txt.
• To prepare for future versions which use 64-bit addressing, made many internal changes for compilation as 64-bit applications and to accommodate large shared memory files.
• Extended the Fortran interface to the WM library Fortran so that it has most of what the C interface has. Documented the Fortran interface.

Return to overview