changes.txt

this a image processing program

* New features :
    - Added 'CImg<T>::RGBtoLUT()' and 'CImg<T>::LUTtoRGB()' that convert an RGB image to an indexed image (using a palette).
    - Added 8 bits display mode support in CImgDisplay.
    - Added 'CImg<T>::RGBtoHSV()' and 'CImg<T>::HSVtoRGB()' that handle the conversion from and to HSV color space.
    - Added 'CImg<T>::kth_smallest()' and 'CImg<T>::median()' that find respectively the kth smallest
      element and the median of values in a CImg<T> instance.
    - Added 'CImg<T>::blur_median()' and 'CImg<T>::get_blur_median()' that apply a median filter on an image.
    - Added 'CImg<T>::load_parrec()' able to read PAR-REC (Philips) image file formats.
    - Added 'CImgDisplay::title()' that can be used to change the display window title.
    - Added 'CImg<T>::RGBtoYUV()' and 'CImg<T>::YUVtoRGB()' that convert between RGB and YUV color spaces.
    - Added 'CImg<T>::RGBtoYCbCr8()' and 'CImg<T>::YCbCr8toRGB()' that convert between RGB and YCbCr color spaces (for unsigned char
    pixels).
    - Added 'CImg<T>::has_same*()' where * can be 'X','Y','Z','XY','XYZ','XYZV' to compare image dimensions.
    - Added 'CImg<T>::is_empty()' that returns 'true' is the image is empty.
    - Added 'CImg<T>::load_dicom()' that loads an image in Dicom file format. This function uses the tool 'medcon' that
    has to be installed (http://xmedcon.sourceforge.net/).
    - Added 'CImg<T>::load_yuv()', 'CImgl<T>::load_yuv()' and 'CImg<T>::save_yuv()', 'CImgk<T>::save_yuv()'
     that load and save respectively one frame or the entire movie from a .yuv file.
    - Added 'CImglSubset<T>' which is a class that sub-references an image list 'CImgl<T>'.

*-----------------------------
*-----------------------------
* Changes from 1.0.8 to 1.0.9
*-----------------------------
*-----------------------------

* Bug corrections :
    - Memory leak corrected in 'CImg<T>::load_bmp()'.
    - CImg now supports the MIPS compiler on SGI (Thanks to Matt Hanson who made it possible).
    - Corrected non-global X11 variables in multi-module projects using CImg.h (thanks also to Matt Hanson for
    his precise bug report).
    - In CImg<T>::load_convert() and save_convert(), random filenames are now tested before read/write.
    - Removed unuseful template in 1D CImg<T>::draw_gaussian() function.
    - Recoded CImgDisplay::render() with XPutPixel() again for maximum compability between X11-based architectures.
    - In CImgStats, the variance field was computed actually as the standart deviation. This is now corrected, and the variance
    is really the variance !

* Optimizations / Modifications :
    - CImgDisplay fields mousex and mousey changed to fields mouse_x and mouse_y. This is done to be more coherent
    with other CImgDisplay fields. The inconvenience should be minor, just replace the fields by the right names.
    - CImgDisplay functions window_posx() and window_posy() changed to fields window_x and window_y.
    - Eigenvalue computation with CImg<T>::eigen() and symeigen() sorts the eigenvalues in inverse order from now on.
    The eigenvector matrix is also transposed, to be more close to the standart functions in eigenvalue computation.
    - Renamed macro variable 'cimg_lapack' in 'cimg_use_lapack'.
    - CImg<T>::save() can now handle alpha channel in PNG images (with 4-channels images).
    - Recoded CImg<T>::noise() to handle numerical limits in template types (esp. for unsigned char and char).
    - Recoded CImg<T>::pow() to be faster with usual int powers (0,1,2,3,4).
    - Added conversions functions for all color spaces (R,G,B), (X,Y,Z), (x,y,Y) and (L,a,b).
    - Added a 'pattern' parameter in 'CImg<T>::draw_ellipse()' and 'CImg<T>::draw_circle()', so that
    only the outline of the ellipse/circle can be drawn.
    - Added an ellipse-based selection in CImg<T>::feature_selection(), when parameter 'feature_type' is set to 3.

* New features/functions :
    - Added an 'Image Zoomer' demo in 'CImg_test.cpp'.
    - Added 'CImg<T>::load_jpeg()' and 'CImg<T>::save_jpeg()' to deal natively with JPEG files, through
      the libjpeg library. You don't need ImageMagick's convert anymore to read and save JPEG files.
      Just define the macro 'cimg_use_jpeg before including 'CImg.h', and link your code with the 'jpeg' lib,
       and it should work fine.
    - Added 'CImg<T>::quicksort()' allowing to sort values inside vectors, and getting corresponding permutations.
    - CImgDisplay::move() is now working well on Windows and Unix.
    - Added 'CImg<T>::SVD()' to compute the SVD of general matrices. The inverse function as well
    as the eigenvalue computation (of symmetric matrices) now uses the SVD. It means that LAPACK is
    not necessary anymore in CImg !
    - Added 'CImgDisplay::screen_dimx()' and 'CImgDisplay::screen_dimy()' to get the resolution of the
    current (full) screen.
    - Added a new 10x13 font, and recoded the functions 'CImgl<T>::get_font();'
    - Added 'cimg::dialog()' that allows to open a simple dialog box with a maximum of 6 choices.
    - Added CImgROI<T>::operator=(), so that one can now affect a channel or plane to an image :
    img.ref_plane(2) = img_plane; Very useful !
    - Added CImg<T>::load_png() and save_png() to deal natively with PNG files, through
     the zlib and libpng libraries. You don't need ImageMagick's convert anymore
      to read and save PNG files. Just define the macro 'cimg_use_png' before including
      'CImg.h', and link your code with the 'zlib' and the 'png' lib, and it should work fine.
       Many thanks to Eric Fausett, for this nice piece of code.
    - Added CImg<T>::load_rgb(), load_rgba(), save_rgb() and save_rgba() that can load and save
      Raw color image data (w/ or wo/ alpha channel).
    - Added CImg<T>::blur_anisotropic() that allows to denoise image by anisotropic filtering.
      Smoothing behavior can be even user-defined.
       Look at  " http://www.greyc.ensicaen.fr/~dtschump/greycstoration/ "
      to learn more about the image regularization technique implemented by this function.

   - ... an tons of minor corrections !

*-----------------------------
*-----------------------------
* Changes from 1.0.7 to 1.0.8
*-----------------------------
*-----------------------------

 * Bug corrections :
    - Removed useless button/keyboard reinitialization in CImgDisplay events
    - Corrected bug in CImg<T>::load_dlm() and load_ascii() functions.
    - Corrected compile bug in CImg<T>::resize_halfXY() and CImg<T>::get_resize_halfXY().
    - Corrected and optimized CImg<T>::get_resize().

 * Optimizations / Modifications :
    - Recoded CImg<T>::fill() for faster value filling.
    - Recoded CImg<T>::draw_graph(). IMPORTANT NOTE : ymin and ymax are now directed up to down, instead
      of down to up, in order to be more coherent with the draw_axeXY() function. This means you will
      probably have to swap the values of these two parameters in the function call, to get similar results
      to CImg 1.0.7 !!
    - Recoded CImg<T>::draw_gaussian() using tensors (parameters changed).
    - CImg is now successfully compiled using the option '-pedantic' with g++. Makefile has been modified.
    - CImg compiles with the DJGPP compiler (tested without display capabilities).
    - Added load/save support for 16bits PNM binary images.
    - Added interleaved mode support and template in constructor :
      template<typename t> CImg(const t *const data_buffer,dx,dy,dz,dv,bool interlaced)
    - Recoded parts of CImg<T>::eigen().
    - Optimized buffer copy in XImage, for faster image display under X11 (minor improvements).
    - Added fields in CImgStats, allowing to get the coordinates of the min/max pixels.

 * New features/functions :
    - Added CImg<T>::load_bmp() and save_bmp() to deal with uncompressed BMP formats, without using ImageMagick.
    - Added CImg<T>::vector(), CImg<T>::matrix() and CImg<T>::tensor().
    - Added CImg<T>::scroll() and CImg<T>::get_scroll to be able to scroll images.
    - Added functions CImg<T>::get_FFT() and CImgl<T>::FFT() for Fast Fourier Transform.
    - Added CImgDisplay::move() to be able to move display windows at specific locations.
    - Added several region split in CImg<T>::get_split().
    - Added 'examples/mcf_levelsets.cpp' : Mean curvature flow of a 2D curve, using level sets.
    - Added 'examples/greycstoration.cpp' : New algorithm for image denoising, inpainting and resizing.
      (see http://www.greyc.ensicaen.fr/~dtschump/greycstoration)
    - Added 'examples/wavelet_atrous.cpp' (by R. Peteri) : Wavelet decomposition of a image.
    - Added a Fourier-based filtering demo in 'CImg_test.cpp', allowing interactive frequency filter creation.
    - Added an Image to ASCII converter, based on simple correlation measure. File : 'examples/image2ascii.cpp'.

 * Abandonned features :
    - Removed bump mapped version of the CImg<T>::draw_triangle() function.
    - Removed 'examples/inpainter.cpp' which is a little bit buggy (problem with multiscale).
    - Removed 'CImg<T>::new_display()' and 'CImgl<T>::new_display', which are confusing and useless. Use
      'new CImgDisplay(img)' and 'new CImgDisplay(list)' instead.
    - Removed 'CImg<T>::get_stats()' and 'CImgl<T>::get_stats', which are useless. Use 'CImgStats(img)'
      and 'CImgStats(list)'.

*-----------------------------
*-----------------------------
* Changes from 1.0.6 to 1.0.7
*-----------------------------
*-----------------------------

 - New License : CImg is now distributed under the CeCiLL License, a free software GPL-like, more adapted
   to French laws. Don't be disturbed by this new License, CImg is still an open source free software !
 - Encapsulation of all CImg classes and functions in the 'cimg_library::' namespace. It avoids eventual
   class names collisions with other libraries. For compatility with your old code, you should
   insert 'using namespace cimg_library;', just after the '#include "CImg.h"' line, in your source code.
 - Removed global typedefs for uchar,uint,ushort,ulong. If you used them before, you'll have to redefine them :
   typedef unsigned char uchar;
   typedef unsigned short ushort;
   typedef unsigned int uint;
   typedef unsigned long ulong;
 - Corrected window resize bug on X11 displays.
 - Added PANDORE-4 file format support, with functions 'CImg::load_pandore()' and 'CImg::save_pandore()' (file extension : .pan)
   (PANDORE is a C++ toolkit for image processing, see http://www.greyc.ensicaen.fr/~regis/Pandore (in French))
 - Added macros CImg_3x3_ref(), CImg_5x5_ref(), .. allowing to use special loops cimg_map3x3,.. with references to C arrays and images.
 - Removed 'normalize' parameter in 'CImg::save()'. Saving normalized images can be still performed
   with 'img.normalize(0,255).save("normalized_image.jpg");'. This clarifies the code and doesn't add so much work at all.
 - Renamed 'CImg::get_sprite()' and 'CImg::sprite()' to 'CImg::get_crop()' and 'CImg::crop()'.
 - Changed 'draw_triangle' function with two textures, so that it corresponds to displacement mapping
   (aka 'bump mapping'), instead of two textures drawn with transparencies
   (which can be still made by two successive calls to single-textured 'draw_triangle' functions,
   with correct opacities).
 - Negative opacity parameter in drawing functions is now supported. It allows to ADD shapes colors to the current
   image pixels (see the new 'ShadeBobs demo in the CImg_test.cpp file to see the concept).
 - Added functions 'CImg::draw_axeX', 'CImg::draw_axeY' and 'CImg::draw_axeXY', allowing to trace labelled axes on images.
 - Slightly changes in CImgDisplay : field 'attributes' has been removed, and replaced by more
   comprehensive field names : events, normalize, fullscreen. Resizing capabilities is now removed
   since all windows can be resized by default.
 - 'CImg<>::load_convert()' and 'CImg<>::save_convert()' do not popup a console window anymore on Windows-based OS,
   when creating code with a WinMain() entry.
 - RGB<->BGR Color inversion problem on some old Unix displays has been corrected.
 - Makefile for compilation on Solaris has been updated.
 - Package structure has been slightly modified for more clarity. Renamed also the CVS module to 'CImg' instead of 'CImg_'
 - Added different compilers projects to ease the compilation of the CImg examples, including
   Visual C++ 6.0, Visual.NET2003, Borland Bcc 5.6, Digital Mars Compiler, Dev-Cpp, Intel ICL.
 - Removed the use of TrackMouseEvent mechanism on Windows, removing compilation bugs with VC++6.0 when trying to create an MDI application.
 - Improved keycode detection under X11. Keycodes should now work for any X11-based system.
 - Support for FreeBSD added (thanks to Thierry Thomas).
 - Corrected bug in matrix multiplication.
 - Corrected bugs in CImg<T>::erode() and CImg<T>::dilate() functions (name inversion).
 - Corrected bug in 'CImg<T>::load_convert()' and 'CImg<T>::save_convert()' when trying to load/save images with pathnames containing spaces.
 - Recoded 'inrcast.cpp', and suppressed 'inrcrop.cpp' in the 'examples/' directory (functionnalities of 'inrcrop' are now included in 'inrcast').
 - And some corrections of minor bugs to improve library stability...
 - Improved error messages to ease program debug.
 - Documentation of the API is much more complete and detailled (still in progress...)

*------------------------------
*------------------------------
* Changes from 1.0.5 to 1.0.6 :
*------------------------------
*------------------------------

 - Drawing functions improved and optimized : Added transparency level to all drawing functions.
 - Added function 'draw_arrow()'
 - Added missing boolean operators.
 - Added CImgDisplay resize event, allowing to create flexible resizing windows.
 - Added 'CImg::draw_gaussian()' functions allowing to draw 1d,2d or 3d gaussians on images.
 - Added 'Oriented convolutions' demos in CImg_test.cpp
 - Recoded pde_TschumperleDeriche2D and 3D.
 - Bugs corrected in arithmetics operators.
 - Optimized image display.
 - Recoded 'CImg::feature_selection()'.
 - Recoded 'CImg::get_rotate()'
 - Renamed 'draw_plot()' to 'draw_graph()', and added cubic drawing style.
 - Renamed 'wait_sync()' to 'wait()'.
 - Renamed 'get_correl' to 'get_correlate' (idem for 'convolve')
 - Removed 'get_correl3x3,5x5,...'. Optimization is now a part of the 'get_correlate' function. (idem for 'convolve').
 - Removed 'get_rotate90,180,270'. Optimization is now a part of the 'get_rotate' function.
 - Compilation supported by gcc.3.4.1 (released july 2004).