Gandalf vision and numerical library

File Release Notes and Changelog

Release Name: Gandalf 1.5

Notes: New features in Gandalf 1.5 --------------------------- New image file formats: DPX, Cineon, SGI, Targa with full DPX support including timecodes. New pixel formats: GAN_RGBX RGB padded out to a word boundary GAN_RGBAS RGB with a small alpha channel GAN_YUVX444 YUV 4-4-4 padded out to a word boundary GAN_YUVA444 YUV with alpha GAN_YUV422 YUV 4-2-2 arranged as UYVY New OpenGL texture rendering module with support for all new and old pixel formats, including fragment shader code where applicable. Added auto-detect for image file formats New cubic spline camera type to define detailed image warps.

---

Changes: 2006/3/17 Version 1.5 released ------------------------------ 2005/02/23 Updated GAN_UPPER to GAN_MATRIXPART_UPPER 2005/07/11 Added more image conversions and display options 2006/02/20 Added support for bit array inside image being preallocated 2005/11/23 Added conversion module image/image_convert.[ch] 2005/11/11 Added function to return boolean value indicating if an image has an alpha channel or not 2005/11/02 New source file common/file_io.[ch], with function to copy a file 2005/10/10 Added DPX, Cineon, SGI and Targa image I/O formats 2005/07/08 Added image/openGL_texture.[ch], code for OpenGL texture rendering of Gandalf images 2005/06/30 New image display test program image/image_display_test.[ch] to test displaying different formats 2005/06/30 Added the first set of extra formats for OpenRL 2005/06/30 Added new image formats: GAN_RGBX, /**< RGB padded out to a word boundary */ GAN_RGBAS, /**< RGB with a small alpha channel */ GAN_YUVX444, /**< YUV 4-4-4 padded out to a word boundary */ GAN_YUVA444, /**< YUV with alpha */ GAN_YUV422, /**< YUV 4-2-2 arranged as UYVY */ 2005/06/28 Added blending functions to image/image_composite.[ch] 2005/06/10 Added extra arguments to camera functions allowing previous transformations to be used in new ones 2005/05/24 Added auto-detect for image file formats 2005/04/19 Added cubic B-spline camera 2005/04/12 Added functions in image/image_channel.[ch] to insert a channel into an image and to fix up an image window 2005/2/22 Version 1.4 released ------------------------------ 2005/02/22 Added code to compute eigenvectors and eigenvalues of symmetric matrices using modified CLAPACK source instead of CCMath with is buggy. 2005/02/10 Added 64-bit headers 2005/02/08 Added const declarations to convolution input images 2005/01/26 Added definitions for 32 bit float images 2004/10/18 Added function to compute homography using normalisation 2004/10/18 Fixed 3x3 homography fitting 2004/10/09 Added function to halve size of image 2004/10/05 Adjustments to essential matrix code 2004/08/25 Added critical section wrapper to make error handler thread-safe in Windows. We need to add pthread code for the other platforms. 2004/07/23 Added test function for cameras being identical 2004/07/20 Added 8/16 bit macros 2004/07/14 Added Mac OpenGL header includes. 2004/07/14 image_composite.c: Builds destination image with correct format, type and dimensions if it isn't initially compatible with source image 2004/07/12 Added an extra test to make sure the correct active subwindow was returned when the image height and width are 0. 2004/07/05 Added filling function with mask 2004/06/22 Corrected masking window function 2004/06/06 Added camera tests 2004/06/04 Added support for passing in alpha matte 2004/06/03 Removed gamma value from Gandalf camera 2004/05/23 Changed offset_x and offset_y in Gan_Image to be signed integers, with appropriate changes to oberon and Gandalf code 2004/05/20 Fixed 16 bit convolution 2004/05/20 Fixed convolution with image expansion 2004/05/19 Revised 1D convolution to allow different behaviour at edge of frame 2004/05/18 Added mask argument to composite function 2004/05/14 Added function to copy linked list with its data 2004/05/13 Added channel maximum value 2004/05/13 Added scaling and min/max functions 2004/05/13 Fixed bug in getting maximum value in channel 2004/03/25 Changed argument names for gandalf lists to avoid C++ namespace collision 2004/03/25 Added Angel's new 2D convolution code 2004/03/22 Added missing 64-bit function 2004/03/18 Added const flags where appropriate 2004/03/17 Added gan_image_clear_window() function to clear an image in a specified rectangular region 2004/03/11 Added setting of pix_data_ptr in gan_image_extract(). Added const qualifiers where appropriate. 2004/03/05 Added UNDEFINED entries to enums 2004/02/26 Added cases for Y convolution from Angel 2004/01/01 Fixes for IA64 compatibility from Torsten Werner 2004/04/01 Added function for 9-grouping of bits in a binary image 2004/01/06 Adjusted Mac installation instructions Fixed file closing bug in image IO 2003/12/29 Added Torsten Werner's changes to make system. Fixed 64 bit compatibility problems in endian_io.[ch]. 2003/12/15 Added #ifdef's around Schur call 2003/12/15 Added progress pointer arguments to Canny and line detector calls 2003/12/15 Fixed Testramework Makefile.in file to refer to GLUT 2003/12/15 Added Linux-specific installation notes 2003/10/13 Version 1.3 released ------------------------------ 2003/10/07 Modified 3D Equalizer camera model to return GAN_ERROR_OUTSIDE_RANGE when failing to undistort a point. 2003/10/01 Changed do loop in bit array copy to memcpy() for performance reasons. 2003/09/09 Removed unnecessary calls to glFlush() in OpenGL functions 2003/09/08 Added pixel offset functions in image/pixel.[ch] 2003/09/03 Fixed undistort function in 3D Equalizer camera format 2003/09/01 Added Mac GL headers to vision/corner_disp.c 2003/08/28 If a point to be projected by a camera is out of range of the model in camera_radial_dist[12].[ch], sets the Jacobian to zero. 2003/08/26 Revised local feature maps for feature matching to allow for border around target image 2003/08/26 Split OpenGL configuration to allow OpenGL-based code that doesn't need the glut GUI library to be compiled into Gandalf. This is so that mokey can call the Gandalf OpenGL functions, e.g. in the camera lens distortion routines, without forcing glut to be present as well. So in the gandalf/config.h file you should have HAVE_OPENGL #defined but not HAVE_GLUT 2003/08/24 Fixed bugs in feature detectors 2003/08/22 Removed flush 2003/08/11 Corrected AND-NOT bit array operation 2003/08/08 Added not-and bit array functions 2003/08/08 Fixed active window function to work with images having one row or column 2003/08/07 Modified LAPACK search to look for lapack, blas and g2c libraries. 2003/8/06 Adjusted installation procedure to put more into configure options (compiler & linker flags), which are taken out of the config file. 2003/07/22 Changed image/image_composite.[ch] to allow subimage to be composited onto larger image 2003/07/18 Added proper floating point support top OpenGL image display 2003/06/05 Fixed bug calling triangular matrix inverse routine in linalg/mat_triangular.[ch]. 2003/05/19 Added function gan_image_get_pixptr(). 2003/05/07 Added boolean cases for grey <--> grey pixel conversion in image/pixel.c 2003/04/04 Fixed bit array insert function 2003/03/30 Added interlace argument to image flip function 2003/03/21 Removed trailing comma from enum definition in vision/camera_defs.h. 2003/03/21 Added image flip routine separately to copy, to allow in-place flip. 2003/03/11 Added transferring of offsets into copied images in convolve1D.c Added reset macro to affine homography fitter. Still doesn't work if the z-coordinates are not 1... 2003/03/03 New compositing routines in image/image_composite.[ch] 2003/02/24 Added 2x2 eigendecomposition of 2x2 symmetric matrices in linalg/2x2matrix_eigsym.[ch] 2003/02/24 Added Hough transform documentation 12-2-2003 Gandalf 1.2 released ------------------------------ 2003/01/31 Added some new ports of CLAPACK/CBLAS code 2002/11/08 Added definitions of powf and sqrtf for platforms without these functions 2002/11/08 Added MAC OS X Project Builder project files and config.h file. 2002/10/06 Added config definition for LIBTOOL to allow for non-default libtool executable. Fixes problems on OS X. Added config.macosx which uses fink libtool 2002/09/30 Added Fast Hough Transform modules to vision package 2002/09/30 Added specific 8/16/32 bit macros 2002/09/30 Added memory stack code common/memory_stack.[ch] to support Fast Hough Transform 2002/07/03 Added complex number structure 2002/06/14 Added some extra matrix/matrix and matrix/vector multiplication functions 2002/06/12 Updated API of feature detectors 2002/06/12 Added outer linear model to radial distortion models 2002/06/12 Removed GIF format. Supporting it seems too fiddly right now. 2002/06/07 Added new files (PBM io) 2002/06/07 Fixed name of gan_symmat22f_sumsqr_s() 2002/06/05 Added array functions (macros) for 8, 16, 32 and 64 bit integers. 2002/06/05 PBM format files 2002/05/27 Removed my calculation of 3x3 matrix eigendecomposition, substituting the CLAPACK version. Changed name of gan_symmat44_eigsym() to gan_symmat44_eigen() to conform to 3x3 version. Added test program for fixed size matrix eigendecomposition. 2002/05/24 Added code to deal with degenerate eigenvalues 2002/05/24 Converted references to row_data in image directory to macro calls. 25-4-2002 Gandalf 1.1 released ------------------------------ 2002/04/22 Finished conversion to Doxygen comment format for documentation 2002/04/19 Added copy functions with flipping: gan_image_copy_flip_q() and gan_image_copy_flip_s(). 2002/04/18 Added Doxygen configuration file doxyfile. Removed ObjectOutline file Gandalf.ool. 2002/04/11 Fixed problem with deleting the only element of a singleton list in common/linked_list.c. Special case for singletons, current_pos is set to -1 and pFirst is set to NULL. 2002/04/08 Added gan_image_get_pixptr_???() functions to return a pointer to a given pixel in an image, to use instead of a direct reference to the row_data field of the image. 2002/03/28 Revision of error handling Fixed png_image_read, fclose() was after return Added code to clean up allocated memory and report memory leaks Fixed memory leaks. Feature detectors now clean their local feature maps 2002/03/27 New install-sh file provided by Matthias Fichtner, to fix Solaris problems Added conversions for signed pixel types Fixed threshold when adding distortion Added gamma initialiser for images 2002/03/26 Added inverse radial distortion model vision/camera[f]_radial_dist1_inv.[ch] 2002/03/26 Added internalize function gan_camera_internalize() for cameras Added new pixel conversions 2002/03/25 Tested configuration file config.sunos for SunOS, using gcc Fixed bug in ..._fill_part() in common/bit_array.c which failed with zero offset 2002/03/22 Added ..._unit_i() macros for fixed size matrices/vectors, which were documented but didn't exist! 2002/03/20 Convolution gan_convolve1D_ui() is now done in doubles instead in floats; overflow bug on line 533 corrected; 2002/03/19 Added new file vision/camera_defs.h of camera definitions Adjusted argument list and meaning of Martin's threshold in image_compare.[ch]. Added RealViz's inverse radial distortion model 2002/03/19 Another major revision of camera stuff. Divided camera projection/ backprojection into modules for each camera model, in the manner of the square matrix modules. 2002/03/15 Added reshape function that actually works Removed incorrect reshape function Removed remaining glutDisplayFunc from image_display.c common/endian_io.[ch] for file I/O independent of little/big endianness Converted binary I/O to little-endian in I/O functions 2002/03/13 Moved array of static strings into header file because of IRIX CC fussiness in bitmap_test.[ch] Added big-endian code to insert routine in bit_array.[ch] Added signed integer definitions to common/misc_defs.h 2002/03/11 Major review of camera code. Projection/backprojection now return errors if pixel is out of range of distortion model. Stereographic and similar models still need to be revised. Split camera.[ch] into five modules. 2002/03/07 New image comparison function in image/image_compare.[ch] 2002/03/06 09:34:50 Added MSDEV Project Gandalf_TestSuite.dsp for Gandalf test suite gandalf/TestFramework Added use of acBuildPathName to generate test input path Added defines for test input/output paths Modified cUnit test program to loop menu until exit is selected 2002/03/05 Added configuration header file config_win32.h for win32 platforms 2002/03/04 Updated to TestFramework/TestInput/TestOutput structure Added ground-truth comparison to bitmap test program Renamed testframework to TestFramework Moved test input files into TestInput directory Removed display_test.c (functionality now in vision_test.c) File DontRemove added to TestOutput directory to make sure that TestOutput directory is not purged by cvs update -P 2002/02/27 Added more tests. The conversion tests in image_test.c now works! Changed image extraction routines to use fast copy routine where possible. Added casts for pixel conversions to fix MSVC warnings. 2002/02/26 Added gamma to camera structure Fixed bugs in GLA header files and added conversions and conversion tests New image inverse utility program utils/invimage.c 2002/02/22 New image inversion function & macros in gandalf/image/image_invert.[ch] 2002/02/20 Added pixel projection/back-projection routines which allow error conditions to be returned in a pointer rather than calling gan_err_register(). Added type definitions GAN_UINT8, GAN_UINT32 etc, #define'd appropriately to the normal types GAN_UCHAR, GAN_USHORT etc. so that we can define unsigned integer variables of specific bit size. Added pixel conversion routine gan_image_pixel_convert() to image/pixel.[ch] which error conditions to be returned in a pointer rather than calling gan_err_register(). 2002/02/15 Added lots of new camera models! 2002/01/24 Added Jacobian output to point projection using a Gandalf camera 2002/01/18 Generalised image pyramid code to allow flexibility in the construction of the low resolution images 2002/01/16 Added data free function to image structure. Added basic JPEG image file read/write and TIFF image file read capability. 2002/01/03 Added GAN_ALL_CHANNELS to convolution specification in image processing algorithms 2002/01/02 Fixed bugs in image display and pixel conversions 2001/12/20 Fixed clean operation to use RM defined symbol and to remove the libraries 2001/12/20 Converted multi-pixel binary image operations to macros when NDEBUG is defined. 2001/12/14 Added mask to feature detectors so that you can find features in a region Added algorithm for Canny edge detection in colour images Added quadratic and cubic equation solvers in common/numerics.[ch] 2-12-2001 Gandalf 1.0 released ------------------------------