Skip to content
Commits on Aug 26, 2010
  1. Whitespace cleanup for sfs.cc

    committed
  2. Made changes to sfs.cc to match Ara's changes. This code exactly matc…

    committed
    …hes the results for one image of the code in Joshbranch. Hell yeah!!
  3. set_default_num_threads fix

    Zack Moratto committed
    If the user was sending a zero to set_default_num_threads, it was
    forgetting to set num_threads back to default amount. Fixed now.
Commits on Aug 25, 2010
  1. Bug fix for PlateCarree Manager

    Zack Moratto committed
    This fixes input images that insert exactly on Plate Carree's
    seam. Also Cylindrical Edge Extension should always be on. Someone was
    overthinking the problem .. but we are edge extending the final
    view. The final view is a global map so cylidrical is an appropriate
    extension.
Commits on Aug 24, 2010
  1. Change error catch to const reference

    Zack Moratto committed
    Doesn't change anything. Yet we do it on principal.
  2. Increased precision in BA report KML.

    Zack Moratto committed
    I need more digits to see what radius those GCPs were being placed.
Commits on Aug 20, 2010
  1. Fix undefined behavior in ConjugateGradient

    Zack Moratto committed
    There was a potential for the threshold to be unachievable. That would
    lead to an infinite loop.
    
    This fix the error where our unit tests would hang on 32bit Linux machines.
Commits on Aug 19, 2010
  1. @novas0x2a

    fix build in non-opencv case and rename to ImageResourceOpenCV

    novas0x2a committed
    The new name fits the DiskImageResource* example slightly better.
  2. @novas0x2a
  3. @novas0x2a
  4. @novas0x2a

    improve bbox gdb printing

    novas0x2a committed
  5. Added F matrix fitting to ipmatch

    Zack Moratto committed
    Works really well for Apollo.
  6. Bug fix plus new error metric

    Zack Moratto committed
    Bug fix for how 8P algorithm gets epipole
    
    Added distance from line error metric.
  7. Added Fundamental Matrix ML Solver

    Zack Moratto committed
    This is the gold standard algorithm described in Multiple View
    Geometry. I followed it the best that I could. However it seems the
    internet community prefers the sampson error over the reprojective
    error. That can happen at a later point.
    
    This algorithm like many others would do great to have a generic sparse LMA.
  8. Switch tests to use FundamentalMatrixErrorFunc for error

    Zack Moratto committed
    Use the functor because it exists. That error functor could possibly
    be improved. I'm not sure if there is any use in running the equations
    forward and backward. I think I've just written the squared error :/
  9. Remove Fundamental Matrix code from Math module

    Zack Moratto committed
    It doesn't make sense. This code is also not my favorite
    implementation .. so I'm just deleting it. The 8p algorithm is already
    available in the camera module. Future Maximium Likelyhood methods for
    F will also require a camera model.
  10. Added Fundamental Matrix 8p to camera geometry

    Zack Moratto committed
    It is a slightly cleaner implementation than the previous.
  11. CameraGeometry refactor

    Zack Moratto committed
    Factored out the similarity normalizer. I'm going to be reusing that
    quite a bit.
    
    I also switch the apply_matrix operation over to a standard
    algorithm. It seems safe to do an in-place modification so that also
    means we dropped an unnecessary copy.
  12. Added progress bar to cnet_editor

    Zack Moratto committed
  13. Warning fix for grassfirealpha

    Zachary Moratto committed
    Made implicit casting into explicit casting. The casting is required
    unfortunately.
Commits on Aug 18, 2010
Commits on Aug 17, 2010
Commits on Aug 16, 2010
  1. Added support for filenames sfsDEMFilename and errorHeightFilename. A…

    committed
    …lso added a text file for debugging purposes to ShapeFromShading.cc.
Commits on Aug 15, 2010
  1. Added a mipmap stopping level option

    Zack Moratto committed
    It is for me to hijack when doing parallel mipmapping.
  2. Factor out triangulation of Control Point

    Zack Moratto committed
    It does make the code easier to read ... but really I'm just doing
    that so ISIS adjust can use it.
Commits on Aug 13, 2010
  1. @anefian

    fixed several bugs in shape from shading from multiple images

    anefian committed
    re-wrote stage 1, 2 and 3 using separate functions.
    removed single image Lambert reflectance model approach.
    computed the Jacobian for points within each image block using also the extended image block;
    the extended image block contains an extra line and column at the right and bottom of each
    image block;
    fixed the bugs in computeRecDerivative by switching in correct order LEFT and TOP points;
  2. Remove f(void) and replace with f()

    Zack Moratto committed
    Apparently Bjarne Stroustrup called this an abomination. Such hard
    feelings against former C programmers. :/
  3. Added a PixelMask interpolation unit test

    Zack Moratto committed
    It makes sure it operates as we expect it.
Commits on Aug 12, 2010
  1. Partial revert aefd201

    Zack Moratto committed
    Thank goodness for unit tests. It really saved me here. I made an
    incorrect assumption in Camera Relations. Point ID is in fact used by
    JFeatures. To set an arbitrary camera ID for IPFeature there is an
    optional 3rd argument now.
    
    This feature is required by an upcoming tool for CNET merging.
  2. @khusmann @zmoratto

    Bugfix: const correctness

    khusmann committed with zmoratto
  3. @khusmann @zmoratto
  4. Make IPFeature use given Camera ID

    Zack Moratto committed
    Previously it was using the information available in Control Measure. Yet not always do I want to use this value.
Something went wrong with that request. Please try again.