[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[libcvd-members] [SCM] libcvd annotated tag, RELEASE_20120202, created.

From: Edward Rosten
Subject: [libcvd-members] [SCM] libcvd annotated tag, RELEASE_20120202, created. RELEASE_20120202
Date: Thu, 02 Feb 2012 16:34:36 +0000

This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "libcvd".

The annotated tag, RELEASE_20120202 has been created
        at  c23029b3b8cfe5a0fee7ea89d206dcfcc8069a04 (tag)
   tagging  ec41afecd7e7f2c80924b55ad56a24cdbd981e7e (commit)
 tagged by  Edward Rosten
        on  Thu Feb 2 17:22:34 2012 +0100

- Log -----------------------------------------------------------------
Tag for release on Feb 2, 2012.

Features added:

Added arctan camera
Added GLES heplers
Allow Zero sized images.
Make Images more STL like
NEON optimizations for halfSample and rgb_to_grey
Sample iOS application
System for generating VCPROJ build files
Efficient Second-order Minimization.
Chainable deinterlacebuffer in open_video_source

Important bugfixes: are:

Many misc bug fixes.
Bug fix in Bayer conversion
Keep up to date with libraries, platforms and compilers.
Thread safety for threading constructs
Removed timer constructs which reduced precision

Cleanup of image I/O code.
Cleanup of build system to remove many #ifdefs
Moved from CVS to git

Dmitri Nikulin (19):
      Correct doxygen in twoThirdsSample()
      Fix a minor doxygen error in vision.h
      Keep separate pthread mutex attributes in Synchronized
      Do not lock in Synchronized destructor
      Create cvd/nocopy.h
      Use NoCopy for Synchronized
      Make Lock in synchronized.h safer
      Remove impossible return in ColourspaceBuffer.seek_to()
      Make colour class constructors safer and more explicit
      Update and cvd headers to newer v4l2
      Merge changes by Dmitri Nikulin
      Regenerate 'configure' after merge
      Fix indentation in files affected by merge
      Extend kernel-video1394.h version checks to Linux 3.x
      Correctly use std::ptrdiff_t
      Extend version checks to Linux 3.x
      Create .gitignore at root and in progs/
      Add /autom4te.cache/ to .gitignore
      Add for installing prereqs

E. Rosten (7):
      Remove misuse of auto_ptr and replace it with vector
      Fix fixing of flipVertical.
      define CVD_IMAGE_DEBUG_INITIALIZE_RANDOM to initialise fresh images with 
random noise
      Remove use of internal macro for YUV420P conditional compilation.
      Ignore vim working space and CVD's compiled programs.
      Shuffle FAST files around, so architecture specific ones exist
      Merge ../wtf/libcvd-git

Ed Rosten (1):
      Compile fixes for MingW including the following:

Edward Rosten (11):
      Updated VS 2008 build system.
      Added simple deinterlacing frame doubling mode into deinterlacebuffer
      Patch from  Achtelik  Markus  to fix compile error on 32bit
      Remove global instance of timer
      Make all the previous changes actually compile.
      Really remove global timer.
      Move implementation of threaded part to source.
      Remove use of global timer.
      Remove global timer in v4l buffer.
      New timer code with cleaner platform separation.
      Reinsert global timer use

Gerhard Reitmayr (1):
      small problems found by clang

Your Name (2):
      Fix dependency resolution for halfSample. Not sure how it ever worked
      Remove strange internal header file which has never served any

aardvaark (27):
      Changed to new TooN (namsepaced) version
      Added missing CVD config header file
      Added Athlon64 detection and optimisation flags
      Fixed CPPLAGS -> CPPFLAGS typo in TooN checking
      Updated to compile with most recent ffmpeg (as well as maintaining 
backward compatability). The reporting of frame rates has changed, as has the 
storage of codecs inside AVStream. Added preprocessor statements to check the 
      Modified to detect and add processor flags for 64-bit (EM64T) Pentium 4s. 
Now generically detects 64-bit processors and sets a CVD_HAVE_64BIT flag.
      Changes to make compile with gcc > 3.3. Added boost-style static assert 
where needed, specified the base class, and removed ebx from list of clobbered 
      Fixed documentation bugs
      Removed %s from LaTeX in Doxygen comments. Doxygen removes the newlines, 
which means that if you use %s, the rest of the LaTeX on the lines following 
gets commented out, confusing the LaTeX compiler!
      Corrected documentation
      Fixed bug in fast_nonmax and added Doxygen documentation.
      Fixed the bug properly this time!
      Added fast_7 and fast_8
      Removed unused parameters to avoid compile warnings
      Minor changes
      Minor changes to documentation, and added to correct Doxygen group
      Documented iterators and begin() and end() in BasicImage, and modified 
documentation for both BasicImage and Image to point at the ImageIO functions 
as well.
      Now with correct function names
      Added unary minus to ImageRef
      Added swap_buffers() to support double buffering
      Minor change to mag_squared(), to make it a little less likely to 
      Added pixel traits for bool, to allow I/O of images of bools
      Changed ir_rounded to correctly round negative numbers (i.e. -1.4 rounds 
now to -1 and -1.5 to -2)
      Changed the semantics of operator< so that an imageref is now less than 
another one if it occurs earlier in the normal scan-line order, i.e. has a 
smaller y-value or, if on the same scanline, a smaller x-value.
      Changed the f2c/g2c Fortran check to one that uses the standard 
AC_FC_LIBRARY_LDFLAGS macro to work out how to link Fortran with C++. This 
makes libcvd now compile with blas/lapack on SuSE 10.x boxes, and doesn't 
appear to break anything else. A dummy MAIN__ does not seem to be needed 
anywhere when using this new macro, but if it should need checking for, the 
AC_FC_DUMMY_MAIN could be used (with associated #defines placed in any 
programs, which is why I've not bothered with it!)
      Fixed documentation typo
      Added test and compiler flags for Intel Core2 processors

barrera (1):
      Changes on to compile on a macosx.

ccs36 (13):
      Changed function name from 'convert_image' to 'convert_image_pair' to 
resolve compile-time conflict between this function and a different function 
with the same signature (but different return type).
      Added specialisation of ScalarConvert to fix incorrect image conversion 
from Rgb<byte> to float.
      Fixed several function calls to match recent changes to utility.h.
      RawVideoFileBuffer is now a templated class. This allows the 
VideoFileBuffer to correctly return frames of type Rgb<byte>.
      This file is now deprecated. However, cleaned out previously committed 
debugging crud so the source can be used again if required.
      Rewritten changes to VideoFileBuffer to fix the bug that caused Rgb<byte> 
images to be deallocated as images of <byte>. The bulk of the code is back in 
the .cc file, which is now un-deprecated. Thanks Ed and Ethan.
      Ethan's fixes to my previous changes to this file. The changed function 
calls should not have template parameters.
      ISO C++ forbids variable-size arrays
      And the necessary delete[]s for the previous change.
      Modified constructors to allow user to specify either bpp, title, or 
      Previous seek_to would only work when seeking to keyframes. This version 
correctly seeks to all frames.
      Removed rogue debugging statement.
      Inserted a check in seek_to() to throw exception if user tries to seek 
past the end of the video stream.

chriskemp (1):
      Opteron changes

ck231 (9):
      Constructor with prior
      Tidy up wls more
      Fixed Ed's typo
      Function to add a prior for a non-zero position
      A ringbuffer class
      Make localvideoframe more CK friendly
      Field changes
      Fix v4l2buffer to use non-blocking (v4l api has changed)
      v4l2buffer - option to select input and chew now works

eads (6):
      Added header files for the Python header files.
      Added python directory where the python interface to libcvd will be 
      Added some initial files for the libcvd python interface.
      Added value_type member typedef to Image, BasicImage, and SubImage 
classes to make them more consistent with STL containers.
      Added a more generic version of drawShape.
      Adapted Image constructors, Image::dup_from, and Image::resize(ImageRef) 
so that images can be initialised or resized with a size of a zero pixel image, 
e.g. size=(0,0), size=(0,k), or size=(k,0) where k is non-zero. Previously, (1) 
CVD::resize(ImageRef(0,0)) caused a segfault, and (2) sizes with one zero 
dimension and the other non-zero was impossible. This has been fixed.

edrosten (464):
      Added support for MMXEXT (restricted subset of SSE which adds useful MMX
      TooN fixes.
      Forgot to add this earlier for -march flags
      Replaced "s with <>s
      Fix #include of byte.h
      Removed experimental #inclide
      Added CVD::Internal::is_POD structure.
      removed POD trait
      Added -without-simd
      Produce position independent code for library (required for Linux on 
      Aligned memory class. Uses Ethan's log N placement delete. Don't use this
      Corrected for TooN changes.
      More TooN fixes
      Remove use of auto_ptr, since it does not use delete[]
      Fixed really silly bug in C++ compiler version checking for gcc.
      Fix type in SGI configuration.
      Fixed SGI compiling bug.
      Added FAST corner detection code
      Compile in FAST.
      Added ImageRef constructor for videodisplay. You can now do
      Put code in the CVD namespace.
      Fixed some bugs which made it miss many corners. Please, don't look at 
the diff.
      image input now detects errors in the istream correctly. Also, there is 
      Added pos (pointer->imageref) method to Image.
      Fix stupid error in last checkin.
      This file is meant to contain portable constructs for CPU specific things
      Fixed bugs in installation.
      Removed some debugging output.
      Added saving images to postscript and EPS
      Added saving images to postscript and EPS
      Fixed incorrect argument types in output_eps_header
      Removed some debugging information
      Remove showpage from the EPS output. It does not seem necessary, and 
makes gv
      Added comments about saving to PS.
      Changes to documentation.
      Added float_type trait.
      Append CVD_ to beginning of all macros.
      Added an interpolation class.
      Fixed bicubic interpolation
      Added documentation.
      Fixed missing TooN::
      Added missing TooN:: (again)
      operator >> can now read what << puts out.
      Fixed docs.
      Fixed bugs relating to configuration and building.
      Removed asserts, replaced Image with BasicImage
      Fixed configuration bug.
      Added array types.
      Tided up documentation.
      Fixed silly compile bug.
      Prevent LaTeX code in comments from producing compiler warnings.
      Changed internals so that image_interpolate can be stored in a container
      Random numbers for different distributions live in here.
      Fixed dumb bugs which should never have been checked in in the first 
      Fixed for x86_64 with respect to i686 asm.
      Removed docs from make install.
      2 phase lookup fixes.
      remove installing of docs.
      Fixed some se3/so3 related bits.
      Fixed some bugs.
      Fixed silly typo
      Removed in class initialization of floats.
      Re-fixed instation of cam_type<...>::fps.
      Update to TooN checking.
      Reordered configure, so that platform specific stuff comes first.
      Fix for gcc 4
      Fix to add -mmmx -msse -msse2, etc where needed. I'm not sure why
      Fix for gcc40
      All S[EO]3 code is now in TooN. Backwards compatability headers will 
exist for
      Removed WLS code from CVD to TooN. Backwards compatibility header will
      Renamed nonmax to fast_nonmax since it relates to FAST.
      Added saving images to a named file (with exceptions)
      Fixed borked jacobian for quintic cameras whereby the principal point 
gets the wrong derivatives.
      Misc fixes to make it compile.
      Added tree based 9,10,11 and 12 point FAST feature detection.
      Fixed more quintic camera jacobian bugs.
      Added support for -DCVD_DISABLE_* for all detected CVD_HAVE_* features.
      Moved pnm_src/*.h to cvd/internal/io/. See old versions for revision
      This file was missing.
      Added missing #include
      Moved files to cvd/internal/io
      Forgot to cvs add this earlier.
      Reinserted code to allow loading to BasicImage types. Exceptions are 
      Removed obsolete file in pnm_src
      Updated documentation
      Made the 16 pixel offsets for FAST accessible through the fast_pixel_ring
      Make it 64 (well N, N > 32) bit clean.
      Added FAST edge detection.
      Added fast_score (max threshold for which feature exists) and updated 
      Fixed fast_score for N <= 8.
      added get_time_of_day which returns seconds since the epoch as a double.
      Added string_to_image_type helper function
      Fixes to string_to_image_type
      Updated docs
      Added proper check for libGL.
      Removed incorrect "static"
      Some fixes for libtiff with static libraries.
      Support old (but still common) libtiff
      Optimization flags for opteron_64
      Solaris fixes.
      Fixed compile buf ig libjpeg is missing
      Work around a piece of aggressive Solarisness.
      OK ln on solaris is more broken than I realised...
      Added enableFPE and tests since not everything is C99 yet!
      Cosmetic fixes.
      Documentation fixes
      Documentation fixes.
      Documentation fixes
      Fix deduction of optimization flags
      Documentation fix
      Added some proper error handling.
      Removed slightly odd line.
      Rewrote TIFF checking. Why do old systems love to do silly things to 
      Unbreak TIFF configuration
      Fixed new[]/delete misatch.
      Fixed "off-the-end" bug in fast_nonmax
      Removed spurious struct foo.
      Fix int (narrow) to int (wide) conversion to use the equivalent to fixed
      Fix CIE conversion so that rgb<byte>(x,x,x) -> byte(x). Done by replacing
      Minor name change.
      Alternate form for glReadPixels
      Added "make ChangeLog"
      Make make ChangeLog work properly.
      string_to_image_type  now returns Unknown, but img_save still defaults
      Make bad image types degenerate to PNM
      Fixed really silly bug.
      Fixed an even sillier bug.
      Added preliminary subimage support.
      Reverting file versions. I meant to check them in on a branch. Ooops.
      Configure check for threads now attempts to append -pthread to the 
      Merging subimage code back in to the main trunk.
      Modified fortran library testing. Unfortunately, AC_FC_LIBRARY_LDFLAGS
      Fixed off by one bug.
      Make aligned mem thread safe (provided the code is compiled with -pthread)
      Now uses CVD::Synchronized.
      Fixed compile error with subimage iterators.
      Misc fixed for (old) FreeBSD:
      Aibned mem now compiles if pthreads are absent.
      FreeBSD fixes: GL is in the X path, not /usr, so this is now added before
      Fixed interpolation to return floating point type.
      halfSample which returns the halfsampled image.
      Removed a really silly cast.
      Conse correct subimage?
      Image loading convinience functions.
      PNG reading support.
      Fix really incredibly stupid bug in that stopped successful reads
      PNG write support.
      Replace std::vector with Internal::simple_vector, since std::vector does
      Fix PNG loading for non specialized types.
      Minor update.
      Removed obsolete error message.
      Now compiles in the presence of missing libraries.
      Added tutorial.
      resize method accepts an optional fill value.
      Check back in pair constructor for images. I didn't mean to remove it.
      Bounds checking for sub_image creation in debug mode.
      Check in a bracket which I seem to have misplaced...
      Fix breakage with subimage iterators and CVD_IMAGE_DEBUG
      For some reason, glob returns a single empty string for an empty glob.
      glClearColor helper to match glColor
      Moved posix_memalign in to compatibility options (ie it should be present)
      A const SubImage can resurn a const sub_image
      Bounds checking for non-const sub_image()
      Bilinear interpolation now does not attempt to access outside the image 
      Lazy image_convert function. Can now do:
      Resize now does not perform allocation if the image has a refcount of 1
      Swapped over tests for pthreads and OpenGL, since on some systems, OpenGL
      Added expose event
      XLookupKeySym doesn't deal with modifiers, but XLookupString does.
      Redone SIMD testing.
      --with-* or --without-* for mmx mmxext sse sse2
      P4 D and --with-cpu=
      Fixed configure script for SIMD.
      Merged new FAST corner detection interface in to the mainline. Your code
      Fix some warnings/errors from -Wextra -pedantic
      Name fix.
      Preliminary code for preventing compilation of some FAST detectors.
      Potential fix to faster_corner_10.
      Fixed some warnings
      Fixed dependencies bug.
      --without-lapack, and "missing programs" list.
      distclean needs to remove .deps
      Fixed a bug in missing programs list. Some cleanup in options display
      Minor fix for --with-cpu
      Minor fix. Seems to help some builds of ffmpeg.
      Maybe fixed bug in faster_corner_9 which was previously in 
      Fix compile error.
      Fixed build dependency bug.
      g++ 3.2.3 compatibility fix.
      Better check for MMX, MMXEXT
      Disable SSE FAST by default.
      Fixed --enable-ssefast bug
      FAST test rig.
      Fixed "missing inverted corners" bug. Bugs still present.
      Fixed off-then edge buf in 9 and 10. All detectors are vlagrind and efence
      SSE fast-9 and SSE fast-10 seem to be working, so they are compiled by 
      Fixed most of the problems in FAST-12 (is_corner_12 takes pointer, width
      SSE fast-12 probably fixed. Can be disabled with --disable-ssefast12
      Minor fix for gcc3, from S. Brumby.
      Fix endianess in 16 bit image writing.
      Changed definitions to match prototypes.
      Removed some compile warnings.
      Fixed compile error for loading Rgb?<unsigned short>
      Strict version of nonmax suppression: points have to be larger (rather
      Function to compute an integral image.
      Fix documentation bug.
      Missing header.
      Debugging version of libCVD:
      Fixed a bug.
      Fixed to prevent rebuilding of debug objects every time.
      Dense tensor voting on gradients for detecting salient edges (for some
      Minor cosmetic/typographic fixes.
      Experimental SSE version.
      Obvious optimization for tensor voting. Used 70% of the time.
      Fix to JPEG reading. Apparently it's hard to tell where the end of a JPEG
      Required include missing if all optional video sources are missing.
      Fixed Make bug. Make mistakenly included fast corner detectors with if
      Fixed borken install if no programs are compiled.
      Cygwin seems to work.
      Make globlist optional, for compatibility with MinGW.
      Make clean now removes debug libraries.
      Retooled the build system:
      Some build fixes
      Mild refactoring. Fixed uninitialized bug from valgrind.
      Add include prefix to include path
      Fix makefile bug
      Fixed a warning
      Added missing file.
      Fixed some dependencies
      Fix jpeg test
      Removed broken warning for very old FFMPEG version
      New config.blah files. Now works with Blackfin CPUs
      Updated docs
      Update docs
      Update docs (again)
      More doc bits
      Build fixes
      Tensor voting now includes the edges
      yuv420p mode for v4lbffer
      Add buffer querying
      Allow operator= on basicimages
      Fixed valgrind error (new[], delete)
      Fix silly error.
      Fix silly error. Again.
      Back port to 2.4 kernel
      Added size option to V4L1Buffer.
      Fix logic for v4lbuffer
      Removed misc. warnings
      Proper shutdown of v4l1buffer
      Fix msssing header for gcc 4.3.0
      Fix small warning.
      More fixes and warning removals for gcc 4.3.0
      Fix future error on depracated feature
      Added doxygen back in to config script
      Fix .fill() for subimages.
      Fix .bmp loading and saving on 64 bit platforms
      Removed some debugging stuff
      Fixed dependency on V4LBuffer
      Added two classes:
      Fix for older compilers, fixes for non x86.
      Added FIR only version of convolveGaussian.
      Added double types.
      Test related to old TooN/new gcc
      Remove spurious exit.
      Removed some warnings
      Rewritten TIFF loading to support float images, amongst other things.
      Added bool support to TIFF.
      Added some test images. Currently float is missing. This will be added
      Fixed TIFF bool loading.
      Removed some debugging
      TIFF saving code. It compiles cleanly, but has not been tested or hooked 
      TIFF saving code.
      TIFF I/O almost works. It compiles cleanly, but there is an error with the
      Floating point TIFF loading and saving now works!  Bool tiff loading and 
      Make the new IO system work on SubImages. The interface now reads 1 line 
      Convert PNG reading to use the new mechanism
      SubImage fixes
      Convert PNG writing to the new system.
      Converted PNM writing to the new system.
      Test for BMP
      Documentation update
      Toonify dome of draw.h
      MingW fixes.
      Support for new FFMPEG (tested against the latest SVN version as of the
      Doc update
      Bug fix for vertical edges.
      Fixed rather silly allocation bug. Interestingly, it's taken this long for
      Added getDisc helper function.
      Fixed same silly allocation bug as earlier
      Added v4l1buffer to videosource.
      Added a new video source for http video streams from some ethernet 
      Fix calibrate for ServerPushJpegStream:
      Added readahead option for jpegstream in videosource.
      Harris camera model. Not fully tested.
      Modified calibrate to generalize the camera model.
      Make it easier to switch cameras.
      Revert calibrate to default to Quintic cameras
      Square Harris camera model. This kind of camera model is for when the 
      Fix for big endian
      Fix for old libtiff.
      Fix compiler error on OSX PPC: fix definition to match header.
      Fix dependencies for video_play_source
      Update to all_options
      GCC 4.3 fix from Casey Link
      Slightly finer grained make install for faster debugging.
      Interpolation for 1D and 2D local maxima.
      Added -mono option
      Fixed mad return.
      Optional verbosity for v4lbuffer, for debugging v4l problems.
      More verbosity
      Fix for bttv driver in v4lbuffer and (depracated?) v4l2buffer.
      Better changelog creator.
      Initial implementation of Live/Flushable/NotLive semantics for video
      Makes a live stream flushable.
      Fixed a warning
      Added a new runtime exception, where a video buffer is unable to grab
      Fix missing header for gcc.
      Stop make .deps  from producing loads of MMX/SSE errors.
      Fixed frame_pending() for VideoBuffer.
      Missing header fix from  Albert Huang
      Removed calls to mcount (used by the profiler), temporarily.
      Remove line count from commit line, since it's variable.
      Minor tweak
      Add in connected components.
      Make connected components actually compile...
      Wrap ifdef around V4L2_CAP_VIDEO_OUTPUT_OVERLAY flag for compatibility
      Removed unneeded #include
      DESTDIR patch from  Albert Huang
      Make types public.
      Added simplistic support for FITS reading. 1, 2 and 3D images are 
      Added simplistic support for FITS reading. 1, 2 and 3D images are 
      SubImage for image_conert.
      Added .ref() member to SubImage, so that you can pass the result of
      SubImage fix.
      Const fix for DrawBox.
      Support for 16 bit unisigned images using the convention of BZERO=32768
      Added more BasicImage prototypes to fix ColourspaceBuffer.
      CVD can now read big endian TIFFs. They start with MM, not II (like the
      Added -pipe to CXXFLAGS.
      Added loading and saving of .txt files as images, of the sort that MATLAB
      Added missing files.
      Throw write error, rather than read error exceptions if writing fails.
      More broken LAPACK install checks
      Implementation of Harris and Shi-Tomasi corners.
      Test program for finding bugs in connected_components.
      Fixed bug in connected_components where diagonal components are 
      Removed spurious incude.
      Naturally, OSX doesn't likle ELF only -soname options.
      Preincrement on SubImageIterator returns const reference to *this.
      Removed relic.
      fix --without-mmx and so on. Also, added --without-inline_asm
      Fixed checking for videodisplay
      Fix spelling.
      Add fill border function
      Fix previous commit
      Fix aliasing
      Add in correct types so convolveGaussian(double) works.
      Useful drawing function
      Faster version of integral image (2.5 ms, vs. 3 ms).
      Added native type and core2 to  march flags
      Allow GLWindow to take an X11 Display string. Ignored for Win32.
      Nanosecond resolution cvd_timer and gettimeofday() on POSIX systems.
      Added missing file
      Fixed makefile so it doesn't die when some debug directories are missing.
      Make window move straight away.
      Report POSIX RT correctly.
      Removed debugging.
      Fixed fits reading (they're not upside-down any more).
      Added FITS writing to Makefile
      FITS file extensions.
      Check for TooN-2 and fail for TooN-1
      Fix some warnings.
      Removed GNUism for portability.
      Fix for STL without .data() method of vector. Is this standard?
      Change rbx to r10, since GCC doesn't like constraints on rbx any more.
      Fix gcc-4.4 compile error (and remove errant printf)
      ARGB datatype patch from Denis Chekhlov.
      Fix previous commit.
      More reliable recompilation.
      Rewrote SubImageIterator to hold a runtime bool to determine if it
      Added 3x3 -> 2x2 image resampling, with SSE2 optimizations.
      Made const members of ConstSubImageIterator non-const. This doesn't seem
      Added correct scalar conversion to bool (ie != 0). This means that loading
      Backport to older kernels (eg 2.6.18), by removing framerate etc
      Removed C99 ism
      Added in missing files?
      Fixed misnamed file.
      Homogeneous morphology like operations. (erode, dilate, binary median).
      Removed spurious assert
      Fast byte morphology.
      Removed the evil extra_data field from video buffer.
      Move open_video_source to DVBuffer3. Also clean up a bit.
      Doc fix
      tabbing fix
      Colourspace conversion in open_video_source. This is implemented by the
      double buffering.
      Doc update.
      Make dc1394v2 disable dc1394v1 by default, since the libraries conflict
      Introduced untyped base class for all video buffers. This is intended to
      Patch to fix compile error from  Adrien Angeli
      Added thread safe size.
      Move OSX path check to the beginning of the section.
      Added verbosity option to DVBuffer3. Accessible using verbose=1 with
      Added explicit format 7 mode selection, eg:
      Fix a compile error for newer compilers.
      Support for libpng 1.4
      Added Bgrx type:
      Added missing file
      SubImage support
      System for passing parameters to image writers
      Added missing file
      Allow access to DV controls
      Merged OpenGL config into the main line
      More extensive FAST test program.
      Fix bug in TIFF loading, when the cooked interface has to be used.
      Remove debugging statement.
      Missing header fix from hauke strasdat.
      Fix OSX OpenGL?
      Remove unused variable.
      PIMPlify PNG reader.
      Added macro based verbose mode to PNG loading.
      Fix loading of palletted PNG images with transparency.
      Revert new-image changes on the mainline
      Check bug fix and verbose mode into the mainline
      Move PNMI/O code over to the proper image loading system.
      Cleanup CVD programs: allow resizing in video_play_source and
      Move non-templated functions into their own source file.
      Dump TIFF directory in verbose mode.
      Check to see if no format can be found and give a more sensible error
      Compute alignment in glDrawPixels to work around buggy drivers.
      Fix broken assert in debug mode.
      Remove bogus inline statement
      Add in missing glFlush()
      Convert BMP reader to new reader system. Conversion of image reading
      Added missing file
      Fix open_video_source for missing glob.
      Remove another spurious inline qualifier.
      Fix open_video_source for missing glob.
      Fix bug reported by Olaf Christ with regards to FFPMEG support.
      FFMPEG headers require #defines to be defined which affect the standard
      Remove support for ancient (2005 and earlier) versions of ffmpeg.
      Add in support for newer FFMPEG to avoide deprecated functions.
      Remove some warnings.
      Uniform image load/save code now complete.
      Add missing file from the previous commit.
      Possible (untested) fixes to VS build files to include new source files.
      Missing header fix (currently affects Windows 7) from Raimund Leitner
      Rearrange code to remove conditional compilation in halfsample.
      Some hacks for keeping the Windows project files up to date.
      Include useful #defines in Windows config to prevent warnings and macro
      Fix to remove multiple istream::get()'s in a multiple statement because
      Small cleanup for selection of various thread yielding options.
      PIMPLify png writing.
      Remove direct use of png_struct and png_info_struct, so that it compiles
      Put libcvd's win32 implementation of gettimeofday inside the CVD namespace
      Change include guard to have a sensible name which doesn't cause strange 
      Add deinterlacing to open_video_source as a chainable buffer.
      OSX 10.6 fixed from Robert Castle
      Update from Robert Castle.
      Fix compile error.
      Pausing and frame-by frame stepping for video_play_source. Also 

er258 (123):
      Initial revision
      Add working v4l2buffer for 2.6
      Change needed to make v4l2buffer work in 2.6
      Adding img_play helper program
      Prevent JPEG loader from trashing the next image in the stream
      Add img_stream_play program. Usage: cat * | img_play_stream
      Adding SE3 inversion program
      Fix bug in makefile preventing se3_inv being built.
      Added Quintic camera
      Chenged default behaviour to old style RepeatLastFrame for compatibility
      Replaced fatal errors with exceptions.
      Added exception classes
      Put EOF check in image factory
      This is an executable and should not have been here in the first place.
      Added ./configure and better configuration checking
      Added some stuff, and moved other stuff in to "Done"
      Added deinterlacebuffer
      Added missing #include
      Adding player program for the new deinterlacing buffer
      Added jpeg saving.
      Added in profiling support
      Inlined trivial accessor functions
      Added profiling support
      Increased buffer size to something too large
      Added config testing for libjpeg
      Fixes for OSX:
      makefile now includes libjpeg.make, as it should.
      Working shared objects under Linux! Now just use -lcvd
      open() returns -1 on failure, not 0. -1 checked for now.
      open() returns -1 on failure, not 0. -1 checked for now.
      Added localvideobuffer where the frames are managed by the program not the
      Added localvideobuffer (and frame) where the frames are managed by the 
      Added "BadPutFrame()" exception, for when the wrong type of frame is put
      Made VideoFileBuffer templated. Added video_play_bw, using this.
      Made VideoFileBuffer templated.
      Made VideoFileBuffer templated. Made player programs respect this.
      OOps, forgot to add these before the last checkin
      Big rewrite of the internal image mangling code.
      Added convert_image() function using the new mangling code.
      Changed some defaults to be more sensible.
      Forgot to add this as part of an earlier checkin.
      Changed some defaults.
      Added TIFF reading support.
      Fixed grey+alpha to grey conversion bug.
      Return Makefile to optimize by default.
      Fixed test for libtiff, again.
      All videobuffers now have a virtual frame_rate()
      All videobuffers now have a virtual frame_rate()
      Fixed bug where glDrawPixels uses the wrong packing for CVD images.
      *** empty log message ***
      Fixed compile error for compilers which can deduce more about templates 
(before instantiation) than gcc 3.3
      Fixed O2Buffer compile error.
      Fixed archiver for GCC compilation on IRIX.
      Fix compile error after the virtual frame_rate() was added.
      Added a cheap and cheerful image viewer.
      OOps, checked in a broken makefile last time...
      Changed iostream.h to iostream
      Make .so and programs by default.
      Added dependency checking for programs
      Changed build system to use autoconf
      File needed for configuration
      Fixed out of date configure script
      Fixed bug in dependency checking
      Fixed dependency ordering so make -j2 now works.
      Added test for ffmpeg version >= 0.4.9
      Added make docs, and fixed --with-ffmpeg=foo
      Fixed remaking of
      Added lowlevel support for bayer pattern conversion
      Added more support for colourspace conversion and ColourspaceBuffer
      Test programs now use colourspacebuffer
      This should have been edded before for colourspace conversions.
      OOps, forgot to add this file...
      I forgot to add this as well.
      TooN and LAPACK are now optional.
      Replace gawk with awk in
      Configure now looks in /sw on OSX
      Added some code to to make debugging of options easier.
      Fixed broken makefile.
      Fixed broken makefile AGAIN.
      Fixed another bug in the makefile.
      On OSX, -I /usr/X11R6/include now gets put on to the include path.
      Added a bit more debugging info to ./configure (print CPPFLAGS)
      Added preliminary support for -framework vecLib in OSX
      Fixed bugs in support for -framework vecLib in OSX
      Fixed bugs in vecLib support: accidental overwriting of CPPFLAGS and 
LDFLAGS, failure to update options variable.
      Working dynamic library support for OS x
      numerics.h back in gl_helpers.h
      Added checking for g++ version >=3, and also look for g++-4, g++-3 if g++ 
is not suitable.
      v4l2 test looks for V4L2_PIX_FMT_GREY instead of videodev2.h, in order to 
work on 2.4 systems
      Dependency checking now works with awks which do not support regexps in 
RS. Also, $AWK is used instead of awk, so AC_PROG_AWK now does something useful.
      Put in correct dependencies for se3 programs.
      Use of detected awk throughout
      Updated docmentation about video capture and portability
      Added support in configure for debugging, profiling and optimizing.
      Changed template parameter To to T to make Doxygen put it in the correct
      Removed last references to old colourspace conversion code.
      Removed old colourspace conversion code.
      Update generated configure file.
      Detection of x86 CPU type on Linux works properly now.
      Detection of TIFF library versiona -- a recent version is needed.
      Use 15fps by default for dvbuffer.
      Missing #include<unistd.h>
      Fix test for TIFF library.
      Prevent optimization of .C files (those with lots of embedded asm).
      Added -lm
      Removed dependence on GPL'd kernel code. kernel-video1394.h is a rewrite
      Added LGPL license boilerplate to the start of each source and header 
      Added authors file.
      Changed GPL to LGPL
      Put in correct address for the FSF
      Put in correct address for FSF
      Added better TooN finding
      Pay proper attention to CXXFLAGS=... for ./configure
      Fix use of LDFLAGS
      Added checking for SSE and SSE2 instruction sets.
      Added support for conversions to two kinds of image simultaneously.
      Added yuv411 to rgb and Y simultaneously.
      Added yuv411 to rgb and Y simultaneously. Put in check for SSE2 to allow
      Corrected dependencies for make install
      Made .so's install properly with symlinks.
      Prevent autoconf from automatically puttin -O2 in CXXFLAGS (.C files 
should be compiled without optimizations).
      Bug report for RGBA

ethaneade (47):
      Retooled pixel conversion/image conversion/image io functionality.  Added
      fixed save defaults for PNM
      fixed save defaults for PNM
      - Fixed some bugs
      Forgot to add yuv422.cpp
      Added V4LBuffer, which does V4L2 in the general sense.  See v4lbuffer.h.  
      added rgb_to_gray.s
      added ReadAheadVideoBuffer
      added ReadAheadVideoBuffer
      added gradient, gradient threshold, and half sample optimizations
      Regularised all the utility functions.  These perform computations over 
      Implementations of utility specialisations.  See utility.h
      An improved fix to the problem Colin addressed recently.  ScalarConvert is
      aligned_mem now calls the appropriate in-place constructor for non-pod
      Added AverageAbs pixel converter (uses average absolute value of 
      Whitespace changes in pixel_traits.h, use std::vector instead of arrays in
      Added project_vector and unproject_vector to quintic camera, and
      convolveWithBox now allows non-square boxes (pass an ImageRef for hwin
      Fixed color struct (now works).
      Fixed glMultMatrix to do rotation and translation in the correct order.
      Minor fixes regarding std namespace.
      Now checks for CVD_HAVE_EMMINTRIN before using intrinsics.
      Added checks to so that configure checks for:
      Fixed bug when using utility functions with small counts.
      Some subtle alignment fixes.
      Added video source url functionality, somewhat like what was discussed a
      Added new operations square and subtract_square with SIMD specializations
      Use posix_memalign for allocating aligned image blocks, when available.
      Added faster_corner_detect_12, which has the same interface as 
      Added checks for alignment and minimum image size.
      Fixed barrier check to be correct in high contrast images.  Removed the 
only inline assembly in the code, so that all the code is intrinsics now. 
is_corner_12 now lives in a separate file, corner_12.h, and has been generated 
as the series of questions with the least number of questions asked before 
termination for a fixed probability of each one being 'yes'.
      Added faster_corner_detect_9 and faster_corner_detect_10.  These run in 
55% and 40% of the
      Added a couple noise projection methods to Camera::Quintic.
      Added GLWindow.  This is a richer implementation of VideoDisplay that 
      Fixed the destructor so that no glX context is current upon destroying the
      The reset() method of cvd_timer now returns the elapsed time since the
      In faster_corner_detect_9 and _10, three locations too close to the end 
of the image
      Added 3x3 median filtering to vision.h, as:
      Added glLine, which takes two points of any type acceptable by glVertex, 
      Added pixel traits for Vector<N> (if TooN is present).
      Added SSE optimized gaussian convolution for float -> float.  Also added 
      Fixes to previous additions.
      Added implementation of Young - van Vliet recursive Gaussian filters for
      Fixed alignment issues for truncated kernel gaussian convolution (for 
images with stride%4!=0) and slightly improved the numerical precision of the 
Young-van Vliet blurring routine.
      Fixed and improved calibration program.

georgklein (55):
      include <fstream>
      Add sharpness control
      Added glOrtho(ImageRef size) wrapper.
      Change glOrtho offset to .375 (red book recommendations)
      Added operator[]
      Added glRasterPos
      *** empty log message ***
      Fixed TooN include
      Added auto_on_off
      Added fast_nonmax_with_scores
      Added mag_squared convenience function
      Added glTexCoord
      Added ir_rounded
      Added area() = product of x and y
      Added compiler warning to explain breakage when compiled without -msse2
      Highly dubious fix without which my DVBuffer doesn't work (??)
      Guard ImageType::JPEG, which is not always defined
      Added glNormal(Vector<3>)
      fast_nonmax used to discard the first and last corners, and all corners
      Added in_image_with_border
      Stop adding spurious -I with no dir, which breaks the AC_CHECK_LIB for 
libGL later on.
      Ooops, left some debug crud in there. Removed now.
      Changed the default output format of ImageRefs to [x y]. This
      const version of operator[]
      Added ImageRef::dot_times(ImageRef), which does what .* does in matlab.
      Documentation fix wrt template params
      Replaced auto_ptr'd array[] with a static array[]. Using auto_ptr with 
      Allow user-selected FPS for v4lbuffer.
      Minimal documentation
      Fix crash in nonmax suppression
      Another minor memory violation
      Fixed faster_corner_9 and 10 crash due to loop being incremented
      Added support for TooN::SO2 and SE2
      Make is_corner_? actually check if things are a corner instead of 
      Added glMultiTexCoord funcs. They are currently gated on
      QTBuffer lets you query the image format string. (Some camera return 
      Added a settingsdialog option to QTBuffer, so that a settings dialog 
      Removed left-over extern pthread_mutex. I think this no longer exists, and
      Fix annoying compiler warning about multi-line comment (and also remove 
      Replaced some more DVBuffer constructor exit()s with exceptions
      fixed sse3 test
      Port the MMX yuv411 stuff to x86_64 (i.e., changed all memory-related
      Minor bug
      Missing include for size_t
      Added DVBuffer3, which is yet another DC1394 buffer.
      Configure now checks for dvbuffer 3 stuff & header fix
      Changed behaviour of dc1394v2 code so that neither ISO channel nor 
      Patch from Pablo Barrera:
      Fix gaussian convolution for images of vectors.
      Added luminance-alpha pixel type La<T> (similar to Rgba<T>)
      Ooops, fix for previous commit
      And another bug in the last commit! Sorry
      Let images of matrices be blurred
      Turn streaming off when closing v4lbuffer, and fix for rgb_to_y non-asm

gerhard (165):
      updated cvd to use new TooN headers and namespace
      moved some vision algorithms from Ethan's code into libcvd. a new 
headerfile vision.h contains the primary interface. some assembly 
implementations for certain specializations were added as well
      small fixes to use all specializations
      added draw functions to paint into images. updated documentation for 
vision and draw
      fixes for gcc 4.0
      fix for gcc 4.0 error:
      fix for gcc 4.0 error:
      fixes for broken static asserts using gcc 4.0. I took the method from 
boost's static assert.
      fixed traits<T>::max_intensity being a static function. gcc 4.0 produced 
the error:
      more helpers to multiply toon matrices on gl matrix stacks, use translate 
and create frustums from camera parameters
      added < operator to ImageRef to enable it for sets, maps, etc
      added inplace vertical flip for images
      make SE3 specific to avoid picking up a TooN::SE3 included by applications
      avoid compile warning for type conversion
      remove undefined function
      moved convolution code to dedicated header/source files and added newer 
convolution template functions used by sift
      make device handle available
      control for autoexposure of v4l1 devices
      added traits for bayer pixel type to allow saving of raw bayer images
      add translation from ImageRef
      fixed declaration for nonmax to correspond with source
      fixed bayer traits to work with recent changes.
      fixed a couple of gcc 4.0 problems
      added glTexSubImage2D to work similar to glDrawPixels on images
      added QuickTime video buffer implementation for OSX. Only tested with OSX 
10.4.x and Fire-i camera.
      fixed block after seek_to
      added configure check and define for pthread_yield which does not exist 
on Mac OS X
      timestamp corrected for start of process time with CVD timer
      more convenience functions
      added frame_pending() implementation to v4lbuffer
      fixed use of select for frame_pending implementations
      avoid warning
      new timeddiskbuffer to return frames from disk with stored timestamps
      fixed gcc 4 compiler errors
      support slices as parameters as well
      my compiler didn't like the '
      precise test for feenableexcept which does not exist on OSX alltough the 
header file does
      added fixes to signatures taken from Ethan's directory to make it work 
      fixed a typo(?)
      changed sample to work with multi component pixel types
      added SumOfSquares pixel conversion
      added a class to control parameters of v4l2 devices
      added documentation from the checkin log
      function to print GL errors and a version of sample that returns the 
sampled value
      added repeated halfsampling function
      correct a current time given as double for the timer starttime
      make ImageRef::next, prev work for min >= max
      removed additional qualifications
      added missing key_iterator typedef
      more texture support, a simple texture id management facility based on 
      removed texture id management again, because it wasn't such a good idea 
after all
      support for mouse move and mouse wheel in eventsummary
      position and resize glwindow
      renamed new methods to be consistent
      add stencil buffer support to glwindow
      small fixes
      added camera calibration program to progs
      set X class hint to interact nicely with WM
      better cursor handling in event summary
      moved jpeglib include into local namespace to avoid conflicting typedef 
with X header
      no TooN guards necessary for new glLine
      make glTexImage functions work with SubImage
      sorted out annoying compile warnings
      calibrate: render video image with rectangular texture instead of gldraw 
(slow on current Suse 10.3 on intel chipset); corrected help
      enabled open_video_source to work with qtbuffer
      added new show settings flag for QTBuffer to videosource
      added default value for untouched  pixels in output image
      fixing some build issues on OSX, automatically create .deps dependencies 
      calibrate works now under OSX with X11 videodisplay
      fix for last fix
      fixed old defines
      avoid allocation of tmp image, if no resize is necessary (image not 
shared and right size)
      support for yuv422 (besides vuy422) for qtbuffer
      moved remaining fast implementations which were excluded by include 
guards into separate source files and have them compiled by the new configure 
system as required
      fixed assembler code to compile on OSX intel
      fixed rand_g to actually use both values from the transformation
      removed warnings about unused parameter
      test on Mac OSX to link against Accelerate framework for lapack before 
anything else. this appears to be the right way to do it.
      fixed png writer for obsolete datatype (triggered by compiling test 
      new fix for not compiling assembler code on OSX. check if lcoal the 
assembler can actually compile our sources, and do not build them, if not
      added missing debug directories and make sure that debug gets recreated 
after a distclean
      porting libcvd to Win32:
      build system update for memalign changes
      added gl text rendering using tesselated characters. see test/font.cpp 
for how to use them and how they look
      updated VC2005 build files and fixed font data to avoid warnings in VC
      allow all fixed matrix variants in glMultMatrix, for example result of 
      added missing conversion again for OSX video support
      added missing object file and fixed typo in configure script, small tweaks
      missing header for gcc 4.3
      some fixes for window position and size in windows implementation
      removed debug output
      constness fixes for Camera::Linear and near, far plane for glOrtho
      updated QTBuffer to changes in VideoBuffer
      added 1D and 2D Haar transforms
      sample does proper conversion between datatypes now and transform can 
write to different output format
      some fixes and 2D versions that return the interpolated value as well
      small fix for Win32, also disabled one very annoying warning
      haar refactored
      added vuy422 to v4lbuffer and more detailed video format output if 
verbose is set
      documented new working camera type, PointGrey USB cameras implement IIDC 
over usb. libdc1394 v2.1 + libusb 1.0 onwards supports these cameras and they 
work out of the box with DVBuffer3
      enabled DVBuffer3 on OSX to use the PointGrey IIDC over USB cameras - the 
latest repository version of libusb works on OSX enabling the whole chain
      support for Format 7 modes in DVBuffer3, if compiled against dc1394 v2
      fixed event handing in glwindow to change gl viewport if resized through 
      large change to bayer support:
      nicer use of templates
      some documentation of new bayer structures
      forgot to check in test program
      added vc2008 project files
      undef some Carbon macro that clashes with TooN2
      added a copy_from member to subimage, uses simple std:copy
      Make libcvd TooN2 compatible.
      used wrong namespace
      used wrong namespace, again
      fixed calibrate after last checkin
      fixed last checkin (need more coffee before hacking)
      updated CVD for Zeros change
      added power on/off interface to DVBuffer3
      better tabs in gltext
      some minor updates for Win32:
      make sure the filenames are sorted
      added missing source files
      updated vc2008 build system to use install locations, added some 
documentation for that
      added missing project file
      fixed crash, if wrong or no render context was present in destructor
      inline doesn't make sense, if no code to inline is available, and MS 
compiler won't export the functions...
      stencil buffer in glwindow also in Win32
      fixes for recent videobuffer work on osx
      corrected small bayer type confusion
      use subimage
      *** empty log message ***
      fix to project file
      glRect interface
      avoiding near and far on Win32 as they are keywords (happy old DOS times)
      added missing files for empty videosource implementations
      fixed bug in Diskbuffer2
      fixed Taylor expansion in interpolate_extremum_value, factor 0.5 was 
missing on the Hessian term
      moved function definition so that the right overloads are picked up (at 
least an issue on win32)
      commented unused args to avoid warnings
      used htons instead of htobe16 for better compatibility
      htonX for Win32
      changed define to select OSX code
      fix for Win32
      fixes to bayer conversion, no more strange patterns in the output
      small fixes and WIN32 compatibility
      reordered definitions so that ir_rounded exists when used - problem here 
on OSX
      update to build files to support spaces in path names
      small fix in image.h for the IMAGE_ASSERT macros, made sample work on 
      fixed typo
      added gl_helpers for OpenGLES1 and split up gl_types as well. made CVD 
respect TooN::DefaultPrecision in some functions and cameras to make porting to 
mobile phones easier. Let me know, if there are any issues
      everything should work on subimages :)
      corrected new files
      finally fixed it
      added dependency between projects
      NEON halfSample implementation
      added rgb to grey conversion for NEON
      reflect last changes to
      fixes to NEON implementation
      project file for XCode4 for iOS development. This is a first simple file 
with a hardcoded configuration defined in build/config.h.iOS. this config file 
will be copied into cvd/ as the first build step. The project is only for 
device targets as it uses the NEON implementations in cvd_src/NEON
      simple iOS video capture and render application for experimenting
      fixes for Win32 using explicit casting functions
      fix for different tr1 header locations in VS
      fix X11 and OpenGL libs on Mac to avoid pulling in libraries from 
/usr/X11R6 which may conflict with MacPorts
      added warp functions that transforms an image from one camera model into 
another one
      added a generic implementation of ESM including different geometric 
transformations and appearance models

gr281 (7):
      added v4l1buffer implementation. currently only tested with Ethan's USB 
2.0 driver
      added v4l1buffer. currently only tested with Ethan's USB 2.0 driver
      test examples for v4l1 buffer
      bayer works now, added some docs about that.
      added timestamps to video frames.
      replaced assertions with proper exceptions in v4l1
      fixed typo in JPEG define

jamesloxam (5):
      64-bit fixes
      Work around for bug in g++ 4.1.0 which causes the configure script to 
think that there are errors in ffmpeg
      extern "C" wrapper around the FFMpeg code, required for newer versions of 
FFMpeg which have dropped C++ support
      Added option '--with-acml' to the configure script, which compiles CVD 
with the lapack and blas routines contained within the ACML library
      Fixed a couple of typos

jfn20 (4):
      minor fixes restoring dc1394 support for mac os x
      ICE client using Apple's QTKit framework
      objective-C code using the QTKit framework to obtain video frames and 
pass them via C data structures
      header file for video_capture.m

jjneubert (5):
      Made transform in vision faster.  Also commented the code so others could 
use it.
      fixed a few bugs
      fixed small error
      added an inv camera derivative
      *** empty log message ***

ok245 (10):
      added fast, lossless image compression as ImageType::CVD
      - more accurate separation of IsConvertible and PixelByPixelConvertible
      added 16bit bayer support (both little and big endian), added workaround 
for incorrectly reported bayer patterns of some unibrain cameras
      added 'reset=1' option to dc1394 videosource to reset stuck firewire 
      added flipHorizontal()
      allow calculation of derivatives of Camera::Harris without first calling 
      added some const qualifiers
      added two forgotten functions for loading camera calibration data
      added missing copy constructor to Rgba class
      const correctness for image_interpolate

pas1001 (39):
      Added end of buffer behaviour selection and seek_to
      Minor tidying up (and made to take a const string&)
      Made destructor virtual (so that polymorphism works)
      Initial version (untested)
      Changed seek_to to take a double
      Changed seek_to to take a double, and moved enum out into VideoBufferFlags
      Added check for ffmpeg libraries
      Added check and test program for ffmpeg libraries and VideoFileBuffer
      Added access function to read the file name
      Changed BadSeek exception to take a double, now that seek_to does
      First version
      Fixed frames_per_second bug (it was reporting seconds per frame)
      Added glReadPixels
      Fixed RepeatLastFrame behaviour
      Now calls put_frame on the real frame when necessary
      Better tiffio config checking
      swapped order of const inline to avoid compile warnings
      removed t from seek definition to avoid unused variable warning
      Added Doxygen documentation
      Renamed WLS to give it a unique name
      Renamed WLS to give it a unique name and marked this file as deprecated
      Renamed internal namespace and classes to make more like the current CVD 
      Split from glhelpers to make file more readable
      Split some definitions into internal directory for readability
      Updated Doxygen documentation
      Added ignore define
      Added source browsing (reference function x) to documentation
      Updated documentation
      started to add check for ffmpeg 0.4.9
      changed to use img rather than pnm load
      fixed silly HAS vs HAVE error in ImageType def
      Removed unused variable from seek declaration. This avoids a compile 
warning, but creates a doxygen warning!
      Added macro definition to include parts of code only for documentation
      Added documentation for ImageType
      Fixed wrong declaration of no-equals operator
      Changed documentation version number
      Swapped initialisation order in constructors to avoid warnings
      added documentation and placed inside CVD namespace
      fixed silly typename typo

paulmcilroy (2):
      Fixed a seg fault resulting from the use of the ebp register as a 
counter. Refactored
      Added Camera::ArcTan for fish-eye lenses derived from the FOV version 
described in Devernary and Faugeras, Straight lines have to be straight

qpan (3):
      add offset to glOrtho for glWindows so that (0,0) is drawn as the top 
left pixel rather than off the display
      uses assign instead of equals to enable integral image to convert between 
e.g. Rgb<T1> and Rgb<T2>
      Added the SimpleTimer class which provides basic timing capabilities 
including average, max and min times over multiple timing cycles. Uses 
cvd_timer internally

syg21 (8):
      Because this code is l33t
      fixed ~VideoDisplay() in so that the display is destroyed 
      Added open by filename.
      fast corners with non-maximal suppression
      removed comments because it breaks some pnm readers like VW because they 
don't handle pnm headers properly.
      syg21: added deprecation #warnings in se3.h, so3.h, wls.h, wls_cholesky.h
      Guard unusual pixel formats with #ifdef, so that it compiles on older 
      Fixed cvd_src/ compile errors.

tangobravo (1):
      Gerhard pointed out the TooN lapack.h header, so I made the Lapack tests 
check for these functions. Fixes configure caching the tests regardless of 
environment variables.

twd20 (8):
      Added support for more recent versions of ffmpeg >= build 4623 (or so)
      class YC is for YUV422 images (such as can be generated by bttv cards)
      V4L2Buffer now templated as V4L2BufferT<T> with a typedef for backward 
      templated v4l2frames
      templated v4l2buffers
      mods to fix bugs in v4l2buffer
      newer ffmpeg requires libavutil to be linked for libavcodec to work.  
configure scripts
      SubImage merged into BasicImage



reply via email to

[Prev in Thread] Current Thread [Next in Thread]