WikiStart »

OpenCV Change Logs¶

OpenCV Change Logs
- 2.4.2
- 2.4.1
- 2.4.0
- 2.4 beta
  - Common changes
  - New functionality
  - OpenCV on GPU
  - OpenCV4Android
  - Performance
  - Bug fixes
  - Known issues
- 2.3.1
  - OpenCV4Android
  - Other New Functionality and Features
  - Documentation
  - Optimization
  - Bug Fixes
  - Known issues
- 2.3
  - Modifications and Improvements since 2.3rc
- 2.3rc
  - General Modifications and Improvements
  - New Functionality, Features
  - Android port
  - Documentation
  - Samples
  - Optimization
  - Bug Fixes
  - Known Problems/Limitations
- 2.2
  - General Modifications and Improvements
  - New Functionality, Features
  - Documentation, Samples
  - Bug Fixes
  - Known Problems/Limitations
- Previous versions

2.4.2¶

July, 2012

Android package introduces a new service-based distribution model (see Android Release Notes for details).

New keypoint descriptor FREAK has been contributed by EPFL group: Kirell Benzi, Raphael Ortiz, Alexandre Alahi and Pierre Vandergheynst. It's claimed to be superior to ORB and SURF descriptors, yet it's very fast (comparable to ORB). Please, see source://trunk/opencv/samples/cpp/freak_demo.cpp.

Improved face recognizer and excellent tutorial on using it has been added by Philipp Wagner. Check the face recognition tutorial.

opencv2.framework for iOS has been created. You can either download the binary from SourceForge or build it yourself using the simple guide. Also, you may be interested to look at some OpenCV on iOS samples, created by our GSoC 2012 students Eduard and Charu: gsoc2012:source:/ios/trunk and read the slides of the tutorial: gsoc2012:source:/ios/trunk/doc/CVPR2012_OpenCV4IOS_Tutorial.pdf

Another 50 bugs have been fixed since 2.4.1 release.

It's not about the code, but now we have the brand-new user site, opencv.org and the new stack overflow-like site with answers to most popular questions answers.opencv.org

GPU module:
- reimplemented CUDA accelerated gpu::PyrLKOpticalFlow for dense and sparse cases. New implementation up 1.5 - 2 times faster then previous GPU optimized. Updated optical flow samples
- implemented resize with area interpolation. CUDA optimized version for integer matrix types up 30 - 35 faster then not optimized OpenCV implementation and up 7 in worst case of 3 channel floating point matrix.

2.4.1¶

June, 2012

The GPU module now supports CUDA 4.1 and CUDA 4.2 and can be compiled with CUDA 5.0 preview.

Added API for storing OpenCV data structures to text string and reading them back:

 1 //==== storing data ====
 2 FileStorage fs(".xml", FileStorage::WRITE + FileStorage::MEMORY);
 3 fs << "date" << date_string << "mymatrix" << mymatrix;
 4 string buf = fs.releaseAndGetString();
 5 
 6 //==== reading it back ====
 7 FileStorage fs(buf, FileStorage::READ + FileStorage::MEMORY);
 8 fs["date"] >> date_string;
 9 fs["mymatrix"] >> mymatrix;

cv::calcOpticalFlowPyrLK now supports precomputed pyramids as input.

Function signatures in documentation are made consistent with source code.

Restored python wrappers for SURF and MSER.

45 more bugs in our bug tracker have been fixed

2.4.0¶

May, 2012

The major changes since 2.4 beta

OpenCV now provides pretty complete build information via (surprise) cv::getBuildInformation().

reading/writing video via ffmpeg finally works and it's now available on MacOSX too.
note 1: we now demand reasonably fresh versions of ffmpeg/libav with libswscale included.
note 2: if possible, do not read or write more than 1 video simultaneously (even within a single thread) with ffmpeg 0.7.x or earlier versions, since they seem to use some global structures that are destroyed by simultaneously executed codecs. Either build and install a newer ffmpeg (0.10.x is recommended), or serialize your video i/o, or use parallel processes instead of threads.

MOG2 background subtraction by Zoran Zivkovic was optimized using TBB.

The reference manual has been updated to match OpenCV 2.4.0 better (though, not perfectly).

>20 more bugs in our bug tracker have been closed (code.opencv.org/projects/opencv/roadmap).

Asus Xtion is now properly supported for HighGUI. For now, you have to manually specify this device by using VideoCapture(CV_CAP_OPENNI_ASUS) instead of VideoCapture(CV_CAP_OPENNI).

2.4 beta¶

April, 2012

As usual, we created 2.4 branch in our repository (code.opencv.org/svn/opencv/branches/2.4), where we will further stabilize the code. You can check this branch periodically, before as well as after 2.4 release.

Common changes¶

At the age of 12, OpenCV got its own home! code.opencv.org is now the primary site for OpenCV development and opencv.org (to be launched soon) will be the official OpenCV user site.

Some of the old functionality from the modules imgproc, video, calib3d, features2d, objdetect has been moved to legacy.

CMake scripts have been substantially modified. Now it's very easy to add new modules - just put the directory with include, src, doc and test sub-directories to the modules directory, create a very simple CMakeLists.txt and your module will be built as a part of OpenCV. Also, it's possible to exclude certain modules from build (the CMake variables "BUILD_opencv_<modulename>" control that).

New functionality¶

The new very base cv::Algorithm class has been introduced. It's planned to be the base of all the "non-trivial" OpenCV functionality. All Algorithm-based classes have the following features:
- "virtual constructor", i.e. an algorithm instance can be created by name;
- there is a list of available algorithms;
- one can retrieve and set algorithm parameters by name;
- one can save algorithm parameters to XML/YAML file and then load them.

A new ffmpeg wrapper has been created that features multi-threaded decoding, more robust video positioning etc. It's used with ffmpeg starting with 0.7.x versions.

features2d API has been cleaned up. There are no more numerous classes with duplicated functionality. The base classes FeatureDetector and DescriptorExtractor are now derivatives of cv::Algorithm. There is also the base Feature2D, using which you can detect keypoints and compute the descriptors in a single call. This is also more efficient.

SIFT and SURF have been moved to a separate module named nonfree to indicate possible legal issues of using those algorithms in user applications. Also, SIFT performance has been substantially improved (by factor of 3-4x).

The current state-of-art textureless detection algorithm, Line-Mod by S. Hinterstoisser, has been contributed by Patrick Mihelich. See objdetect/objdetect.hpp, class Detector.

3 face recognition algorithms have been contributed by Philipp Wagner. Please, check opencv/contrib/contrib.hpp, FaceRecognizer class, and opencv/samples/cpp/facerec_demo.cpp.

2 algorithms for solving PnP problem have been added. Please, check flags parameter in solvePnP and solvePnPRansac functions.

Enhanced LogPolar implementation (that uses Blind-Spot model) has been contributed by Fabio Solari and Manuela Chessa, see opencv/contrib/contrib.hpp, LogPolar_* classes and opencv/samples/cpp/logpolar_bsm.cpp sample.

A stub module photo has been created to support a quickly growing "computational photography" area. Currently, it only contains inpainting algorithm, moved from imgproc, but it's planned to add much more functionality.

Another module videostab (beta version) has been added that solves a specific yet very important task of video stabilization. The module is under active development. Please, check opencv/samples/cpp/videostab.cpp sample.

findContours can now find contours on a 32-bit integer image of labels (not only on a black-and-white 8-bit image). This is a step towards more convenient connected component analysis.

Canny edge detector can now be run on color images, which results in better edge maps

Python bindings can now be used within python threads, so one can write multi-threaded computer vision applications in Python.

OpenCV on GPU¶

Different Optical Flow algorithms have been added:
- Brox (contributed by NVidia)
- PyrLK - both Dense and Sparse variations
- Farneback

New feature detectors and descriptors:
- GoodFeaturesToTrack
- FAST/ORB which is patent free replacement of SURF.

Overall GPU module enhancements:
- The module now requires CUDA 4.1 or later;
- Improved similarity of results between CPU and GPU;
- Added border extrapolation support for many functions;
- Improved performance.

pyrUp/pyrDown implementations.

Matrix multiplication on GPU (wrapper for the CUBLAS library). This is optional, user need to compile OpenCV with CUBLAS support.

OpenGL back-end has been implemented for highgui module, that allows to display GpuMat directly without downloading them to CPU.

OpenCV4Android¶

See the Android Release Notes.

Performance¶

A few OpenCV functions, like color conversion, morphology, data type conversions, brute-force feature matcher have been optimized using TBB and/or SSE intrinisics.

Along with regression tests, now many OpenCV functions have got performance tests. Now for most modules one can build opencv_perf_<modulename> executables that run various functions from the particular module and produce a XML file. Note that if you want to run those tests, as well as the normal regression tests, you will need to get (a rather big) code.opencv.org/svn/opencv/trunk/opencv_extra directory and set environment variable OPENCV_TEST_DATA_PATH to "<your_copy_of_opencv_extra>/testdata".

Bug fixes¶

In this version we fixed literally hundreds of bugs. Please, check code.opencv.org/projects/opencv/versions/1 for a list of fixed bugs.

Known issues¶

When OpenCV is built statically, dynamically created classes (via Algorithm::create) can fail because linker excludes the "unused" object files. To avoid this problem, create classes explicitly, e.g
```
1 Ptr<DescriptorExtractor> d = new BriefDescriptorExtractor;
```

2.3.1¶

August, 2011

OpenCV4Android¶

OpenCV Java bindings for Android platform are released in ''Beta 2'' quality. A lot of work is done to make them more stable and easier to use. Currently Java API has about 700 different OpenCV functions and covers 8 OpenCV modules including full port of features2d.

Other New Functionality and Features¶

Retina module has been contributed by Alexandre Benoit (in opencv_contrib module). See the new retina sample and https://sites.google.com/site/benoitalexandrevision/.

Support for Ximea cameras (www.ximea.com/) in highgui has been contributed by Ximea developers.

Planar subdivisions construction (Delaunay triangulation and Voronoi tessellation) have been ported to C++. See the new delaunay2.cpp sample.

Several new Python samples have been added.

FLANN in OpenCV has been upgraded to v1.6. Also, added Python bindings for FLANN.

We now support the latest FFMPEG (0.8.x) that features multi-threaded decoding. Reading videos in OpenCV has never been that fast.

Documentation¶

Quite a few new tutorials have been added. Check opencv.itseez.com/trunk for the up-to-date documentation.

Optimization¶

Performance of the sparse Lucas-Kanade optical flow has been greatly improved. On 4-core machine it is now 9x faster than the previous version.

Bug Fixes¶

Over 100 issues have been resolved since 2.3 release. Most of the issues (closed and still open) are listed at code.opencv.org/projects/opencv/issues?set_filter=1.

Known issues¶

TBB debug binaries are missed in the Windows installer. Here is a workaround:
- Download tbb30_20110427oss_win.zip from the TBB website.
- Unzip and copy the tbb*_debug.dll files from bin/<ARCH>/<COMPILER> to the corresponding folder in the installed OpenCV location in <OPENCVCV_ROOT>/build/common/tbb/<ARCH>/<COMPILER>

2.3¶

July, 2011

Modifications and Improvements since 2.3rc¶

A few more bugs reported in the OpenCV bug tracker have been fixed.

Documentation has been improved a lot! The new reference manual combines information for C++ and C interfaces, the OpenCV 1.x-style Python bindings and the new C++-style Python bindings. It has also been thoroughly checked for grammar, style and completeness.
Besides, there are new and updated tutorials.

The up-to-date online documentation is available at opencv.itseez.com.

The new binary package includes various pre-compiled libs:
https://sourceforge.net/projects/opencvlibrary/files/opencv-win/2.3/
Unfortunately, it's not a full-scale installation package, but simply a self-extracting archive with a readme.txt supplied.
The installation package is probably to come in the next version.

VS2005 should build OpenCV 2.3 out of the box, including DirectShow support.

ffmpeg bindings are now available for all Windows users via compiler- and configuration- and
version-independent opencv_ffmpeg.dll (for 32-bit compilers) and opencv_ffmpeg_64.dll (for 64-bit compilers).

2.3rc¶

June, 2011

General Modifications and Improvements¶

Buildbot-based Continuous Integration system is now continuously testing OpenCV snapshots. The status is available at buildbot.itseez.com

OpenCV switched to Google Test (code.google.com/p/googletest/) engine for regression and correctness tests. Each module now has test sub-directory with the tests.

New Functionality, Features¶

Many functions and methods now take InputArray/OutputArray instead of cv::Mat references. It retains compatibility with the existing code and yet brings more natural support for STL vectors and potentially other "foreign" data structures to OpenCV. See opencv.itseez.com/modules/core/doc/intro.html#inputarray-and-outputarray for details.

core
- LAPACK is not used by OpenCV anymore. The change decreased the library footprint and the compile time. We now use our own implementation of Jacobi SVD. SVD performance on small matrices (2x2 to 10x10) has been greatly improved; on larger matrices it is still pretty good. SVD accuracy on poorly-conditioned matrices has also been improved.
- Arithmetic operations now support mixed-type operands and arbitrary number of channels.

features2d
- Completely new patent-free BRIEF and ORB feature descriptors have been added.
- Very fast LSH matcher for BRIEF and ORB descriptors will be added in 2.3.1.

calib3d
- A new calibration pattern, circles grid, has been added. See findCirclesGrid() function and the updated calibration.cpp sample. With the new pattern calibration accuracy is usually much higher.

highgui
- [Windows] videoInput is now a part of highgui. If there are any problems with compiling highgui, set WITH_VIDEOINPUT=OFF in CMake.

stitching
- opencv_stitching is a beta version of new application that makes a panorama out of a set of photos taken from the same point.

python
- Now there are 2 extension modules: cv and cv2. cv2 includes wrappers for OpenCV 2.x functionality. opencv/samples/python2 contain a few samples demonstrating cv2 in use.

contrib
- A new experimental variational stereo correspondence algorithm StereoVar has been added.

gpu
- the module now requires CUDA 4.0 or later; Many improvements and bug fixes have been made.

Android port¶

With support from NVIDIA, OpenCV Android port (which is actually not a separate branch of OpenCV, it's the same code tree with additional build scripts) has been greatly improved, a few demos developed. Camera support has been added as well.
See Android Release Notes for details.

Documentation¶

OpenCV documentation is now written in ReStructured Text and built using Sphinx (sphinx.pocoo.org).

It's not a single reference manual now, it's 4 reference manuals (OpenCV 2.x C++ API, OpenCV 2.x Python API, OpenCV 1.x C API, OpenCV 1.x Python API), the emerging user guide and a set of tutorials for beginners.

Style and grammar of the main reference manual (OpenCV 2.x C++ API) have been thoroughly checked and fixed.

Online up-to-date version of the manual is available at opencv.itseez.com

Samples¶

Several samples using the new Python bindings (cv2 module) have been added: code.opencv.org/svn/opencv/branches/2.3/opencv/samples/python2

Optimization¶

Several ML algorithms have been threaded using TBB.

Bug Fixes¶

Over 250 issues have been resolved. Most of the issues (closed and still open) are listed at code.opencv.org/projects/opencv/issues?set_filter=1.

Known Problems/Limitations¶

Documentation (especially on the new Python bindings) is still being updated. Watch opencv.itseez.com for updates.

Android port does not provide Java interface for OpenCV. It is going to be added to 2.3 branch in a few weeks.

2.2¶

December, 2010

General Modifications and Improvements¶

The library has been reorganized. Instead of cxcore, cv, cvaux, highgui and ml we now have several smaller modules:
- opencv_core - core functionality (basic structures, arithmetics and linear algebra, dft, XML and YAML I/O ...).
- opencv_imgproc - image processing (filter, GaussianBlur, erode, dilate, resize, remap, cvtColor, calcHist etc.)
- opencv_highgui - GUI and image & video I/O
- opencv_ml - statistical machine learning models (SVM, Decision Trees, Boosting etc.)
- opencv_features2d - 2D feature detectors and descriptors (SURF, FAST etc., including the new feature detectors-descriptor-matcher framework)
- opencv_video - motion analysis and object tracking (optical flow, motion templates, background subtraction)
- opencv_objdetect - object detection in images (Haar & LBP face detectors, HOG people detector etc.)
- opencv_calib3d - camera calibration, stereo correspondence and elements of 3D data processing
- opencv_flann - the Fast Library for Approximate Nearest Neighbors (FLANN 1.5) and the OpenCV wrappers
- opencv_contrib - contributed code that is not mature enough
- opencv_legacy - obsolete code, preserved for backward compatibility
- opencv_gpu - acceleration of some OpenCV functionality using CUDA (relatively unstable, yet very actively developed part of OpenCV)

If you detected OpenCV and configured your make scripts using CMake or pkg-config tool, your code will likely build fine without any changes. Otherwise, you will need to modify linker parameters (change the library names) and update the include paths.

It is still possible to use #include <cv.h> etc. but the recommended notation is:

1 #include "opencv2/imgproc/imgproc.hpp" 
2 ...

Please, check the new C and C++ samples (code.opencv.org/svn/opencv/trunk/opencv/samples), which now include the new-style headers.

The new-style wrappers now cover much more of OpenCV 2.x API. The documentation and samples are to be added later. You will need numpy in order to use the extra added functionality.
SWIG-based Python wrappers are not included anymore.

OpenCV can now be built for Android (GSoC 2010 project), thanks to Ethan Rublee; and there are some samples too. Please, check OpenCV4Android

The completely new opencv_gpu acceleration module has been created with support by NVIDIA. See below for details.

New Functionality, Features¶

core
- The new cv::Matx<T, m, n> type for fixed-type fixed-size matrices has been added. Vec<T, n> is now derived from Matx<T, n, 1>. The class can be used for very small matrices, where cv::Mat use implies too much overhead. The operators to convert Matx to Mat and backwards are available.
- cv::Mat and cv::MatND are made the same type: typedef cv::Mat cv::MatND.
  Note that many functions do not check the matrix dimensionality yet, so be careful when processing 3-, 4- ... dimensional matrices using OpenCV.
- Experimental support for Eigen 2.x/3.x is added (WITH_EIGEN2 option in CMake). Again, there are convertors from Eigen2 matrices to cv::Mat and backwards. See modules/core/include/opencv2/core/eigen.hpp.
- cv::Mat can now be print with "<<" operator. See opencv/samples/cpp/cout_mat.cpp.
- cv::exp and cv::log are now much faster thanks to SSE2 optimization.

imgproc
- color conversion functions have been rewritten;
- RGB->Lab & RGB->Luv performance has been noticeably improved. Now the functions assume sRGB input color space (e.g. gamma=2.2). If you want the original linear RGB->L** conversion (i.e. with gamma=1), use CV_LBGR2LAB etc.
- VNG algorithm for Bayer->RGB conversion has been added. It's much slower than the simple interpolation algorithm, but returns significantly more detailed images
- The new flavors of RGB->HSV/HLS conversion functions have been added for 8-bit images. They use the whole 0..255 range for the H channel instead of 0..179. The conversion codes are CV_RGB2HSV_FULL etc.
- special variant of initUndistortRectifyMap for wide-angle cameras has been added: initWideAngleProjMap()

features2d
- the unified framework for keypoint extraction, computing the descriptors and matching them has been introduced. The previously available and some new detectors and descriptors, like SURF, FAST, StarDetector etc. have been wrapped to be used through the framework. The key advantage of the new framework (besides the uniform API for different detectors and descriptors) is that it also provides high-level tools for image matching and textured object detection. Please, see documentation opencv.itseez.com/modules/features2d/doc/common_interfaces_of_feature_detectors.html
  and the C++ samples:
  - descriptor_extractor_matcher.cpp - finding object in a scene using keypoints and their descriptors.
  - generic_descriptor_matcher.cpp - variation of the above sample where the descriptors do not have to be computed explicitly.
  - bagofwords_classification.cpp - example of extending the framework and using it to process data from the VOC databases: pascallin.ecs.soton.ac.uk/challenges/VOC/
- the newest super-fast keypoint descriptor BRIEF by Michael Calonder has been integrated by Ethan Rublee. See the sample opencv/samples/cpp/video_homography.cpp
- SURF keypoint detector has been parallelized using TBB (the patch is by imahon and yvo2m)

objdetect
- LatentSVM object detector, implementing P. Felzenszwalb algorithm, has been contributed by Nizhniy Novgorod State University (NNSU) team. See opencv/samples/c/latentsvmdetect.cpp

calib3d
- The new rational distortion model:
  x' = x * (1 + k₁ * r² + k₂ * r⁴ + k₃ * r⁶)/(1 + k₄ * r² + k₅ * r⁴ + k₆ * r⁶) + <tangential_distortion for x>,
  y' = y * (1 + k₁ * r² + k₂ * r⁴ + k₃ * r⁶)/(1 + k₄ * r² + k₅ * r⁴ + k₆ * r⁶) + <tangential_distortion for y>
  
  has been introduced. It is useful for calibration of cameras with wide-angle lenses.
  Because of the increased number of parameters to optimize you need to supply more data to robustly estimate all of them. Or, simply initialize the distortion vectors with zeros and pass CV_CALIB_RATIONAL_MODEL to enable the new model CV_CALIB_FIX_K3 + CV_CALIB_FIX_K4 + CV_CALIB_FIX_K5 or other such combinations to selectively enable or disable certain coefficients.
- rectification of trinocular camera setup, where all 3 heads are on the same line, is added. see samples/cpp/3calibration.cpp

ml
- Gradient boosting trees model has been contributed by NNSU team.

highgui
- Experimental Qt backend for OpenCV has been added as a result of GSoC 2010 project, completed by Yannick Verdie. The backend has a few extra features, not present in the other backends, like text rendering using TTF fonts, separate "control panel" with sliders, push-buttons, checkboxes and radio buttons, interactive zooming, panning of the images displayed in highgui windows, "save as" etc. Please, check the youtube videos where Yannick demonstrates the new features: www.youtube.com/user/MrFrenchCookie#p/u
- The new API is described here: opencv.itseez.com/modules/highgui/doc/qt_new_functions.html To make use of the new API, you need to have Qt SDK (or libqt4 with development packages) installed on your machine, and build OpenCV with Qt support (pass -DWITH_QT=ON to CMake; watch the output, make sure Qt is used as GUI backend)
- 16-bit and LZW-compressed TIFFs are now supported.
- You can now set the mode for IEEE1394 cameras on Linux.

contrib
- Chamfer matching algorithm has been contributed by Marius Muja, Antonella Cascitelli, Marco Di Stefano and Stefano Fabri. See samples/cpp/chamfer.cpp

gpu
This is completely new part of OpenCV, created with the support by NVIDIA.
Note that the package is at alpha, probably early beta state, so use it with care and check OpenCV SVN for updates.
In order to use it, you need to have the latest NVidia CUDA SDK installed, and build OpenCV with CUDA support (-DWITH_CUDA=ON CMake flag).

All the functionality is put to cv::gpu namespace. The full list of functions and classes can be found at
opencv/modules/gpu/include/opencv2/gpu/gpu.hpp, and here are some major components of the API:
- image arithmetics, filtering operations, morphology, geometrical transformations, histograms
- 3 stereo correspondence algorithms: Block Matching, Belief Propagation and Constant-Space Belief Propagation.
- HOG-based object detector. It runs more than order of magnitude faster than the CPU version!
  See opencv/samples/gpu

python bindings
A lot more of OpenCV 2.x functionality is now covered by Python bindings.
These new wrappers require numpy to be installed
(see opencv.willowgarage.com/wiki/InstallGuide for details).

Likewise the C++ API, in the new Python bindings you do not need to allocate output arrays.
They will be automatically created by the functions.

Here is a micro example:
```
1   import cv
2 
3   a=cv.imread("lena.jpg",0)
4   b=cv.canny(a, 50, 100, apertureSize=3)
5   cv.imshow("test",b)
6   cv.waitKey(0)
```
In the sample a and b are normal numpy arrays, so the whole power of numpy and scipy can now be combined with OpenCV functionality.

Documentation, Samples¶

Links to wiki pages (mostly empty) have been added to each function description, see opencv.willowgarage.com

All the samples have been documented with default output ''(0 or incomplete number of parameters)'' set to print out "howto" run instructions [Gary]; most samples have been converted to C++ to use the new OpenCV API.

Bug Fixes¶

Over 300 issues have been resolved. Most of the issues (closed and still open) are listed at code.opencv.org/projects/opencv/issues?set_filter=1

The old bug tracker at https://sourceforge.net/projects/opencvlibrary/ is now closed for updates. As soon as all the still relevant bug reports will be moved to code.ros.org, the old bug tracker will be completely deleted. Please, use the new tracker from now on.