Skip to content

Latest commit

 

History

History
492 lines (445 loc) · 47.2 KB

CHANGELOG.md

File metadata and controls

492 lines (445 loc) · 47.2 KB
Raw
  • Add more appropriate default pixel formats for JPEG formats in FFmpegFrameRecorder (issue #410)

May 15, 2016 version 1.2

  • Optimize AndroidFrameConverter a bit and add a test (pull #379)
  • Fix DC1394FrameGrabber on the Windows platform (issue bytedeco/procamcalib#4)
  • Support AVPacket in FFmpegFrameGrabber and FFmpegFrameRecorder to copy without re-encoding (issue #93)
  • Lower Maven prerequisite in the pom.xml file to 3.0 (issue bytedeco/javacpp#93)
  • Add new PrincipalComponentAnalysis sample (pull #373)
  • Upgrade OpenCVFrameRecorder to use the new C++ VideoWriter API (pull #370)
  • Upgrade OpenCVFrameGrabber to use the new C++ VideoCapture API (pull #361)
  • Add CaffeGooglenet.java sample for the opencv_dnn module (pull #341)
  • Clean up IPCameraFrameGrabber and fix incorrectly reading some headers (pull #323, pull #345)
  • Fix swallowed InterruptedException and throw appropriate exception in FrameGrabber.start() (issue #315)
  • Fix IPCameraFrameGrabber.stop() not checking for null (pull #300)
  • Upgrade dependencies for OpenCV 3.1.0, FFmpeg 3.0.2, FlyCapture 2.9.3.43, libdc1394 2.2.4
  • Let users call FFmpegFrameFilter.push(null) to indicate EOF, as required by some filters like "palettegen" (issue #287)
  • Call cvHaarDetectObjects() with CV_HAAR_FIND_BIGGEST_OBJECT | CV_HAAR_DO_ROUGH_SEARCH instead of CV_HAAR_DO_CANNY_PRUNING in the face detection samples to get acceptable performance with OpenCV 3.0 (issue #272)
  • Change WakeLock for keepScreenOn in AndroidManifest.xml file and add setPreviewDisplay() call on surfaceChanged() event for the RecordActivity sample (pull #269, pull #271)

October 25, 2015 version 1.1

  • Make FrameConverter for images return null when Frame.image == null (issue #249)
  • Add FFmpegLogCallback to redirect easily to Java log messages from FFmpeg
  • Upgrade all Maven dependencies and plugins to latest versions, thus bumping minimum requirements to Java SE 7, Android 4.0, and Maven 3.0
  • Fix broken FFmpegFrameGrabber.grabImage() after setTimestamp() (issue #236)
  • Add FFmpegFrameGrabber.grabSamples() to grab only audio samples, and ignore video frames (issue #235)
  • Fix broken setVideoCodecName() and setAudioCodecName() for FFmpegFrameRecorder (issue #229)
  • Remove FaceRecognition.java sample, which requires the deprecated opencv_legacy module (issue #200)
  • Fix potential crash in ObjectFinder with FLANN (issue #210)
  • Add FFmpegFrameFilter to let users process Frame images with libavfilter easily (issue #164)
  • Add FaceRecognizerInVideo.java sample that does a combo of face detection and recognition (issue #203)
  • Return AVStream.r_frame_rate when AVStream.avg_frame_rate is invalid in FFmpegFrameGrabber.getFrameRate() (issue #292)
  • Update some samples to make them work with OpenCV 3.0
  • Add new convenience FFmpegFrameRecorder.record(Frame frame, int pixelFormat) method (issue #181)
  • Let Java2DFrameConverter.copy() from ByteBuffer with 4 channels to BufferedImage.TYPE_INT_RGB, among others, also taking into account the flipChannels argument (issue #181)

July 11, 2015 version 1.0

  • Offer the Apache License, Version 2.0, as a new choice of license, in addition to the GPLv2 with Classpath exception
  • Upgrade support to OpenCV 3.0.0
  • Upgrade supported FFmpeg API to the 2.7 release branch
  • Switch descriptor used by ObjectFinder from SURF to AKAZE
  • Let users get resized images from FFmpegFrameGrabber by calling setImageWidth() and setImageHeight() before start()
  • Add check for supported display size in the RecordActivity sample (pull #153)
  • Clarify the semantics of FrameConverter (issue #150)
  • Fix FFmpegFrameRecorder not saving the last few frames, especially when encoding with x264 (issue #50)
  • Add FrameConverterTest and fix a couple of bugs uncovered by it
  • Make Frame implements Indexable for easy and efficient access to image pixels
  • Fix AbstractMethodError thrown from OpenCVFrameConverter on some versions of the JDK (issue #143)
  • Add FFmpegFrameGrabber.grabImage() method to restore the functionality previously provided by IplImage grab() (issue #116)
  • Give users of FFmpegFrameGrabber and FFmpegFrameRecorder access to more options and metadata (issue #132)
  • Add the ability to specify from which video and audio streams FFmpegFrameGrabber should grab from (issue #135)
  • Fix Java2DFrameConverter when used with BufferedImage.TYPE_INT_RGB or other types based on int (issue #140)
  • Add new WebcamAndMicrophoneCapture sample (pull #131)
  • Add aspectRatio property to FrameGrabber and FrameRecorder, to be able to use pixel aspect ratios other than 1.0 (issue #90)

April 4, 2015 version 0.11

  • Upgrade support to OpenCV 2.4.11
  • Upgrade supported FFmpeg API to the 2.6 release branch
  • Add new Square sample, thanks to Geir Ruud
  • Add AndroidFrameConverter, Java2DFrameConverter and OpenCVFrameConverter, and use them to refactor Frame, CanvasFrame, FrameGrabber, and FrameRecorder in a way to help users avoid coupling with Android, Java 2D, or OpenCV (issue #84)
  • Fix Demo class in the README.md file (issue #102)
  • Add new ColoredObjectTrack sample (pull #99)
  • Add option property to FFmpegFrameGrabber to let users set such things as "analyzeduration", "probesize", or "list_devices"
  • Fix "AVFrame.format is not set" and "AVFrame.width or height is not set" warning messages (issue #76)

December 23, 2014 version 0.10

  • Upgrade support to OpenCV 2.4.10
  • Upgrade supported FFmpeg API to the 2.5 release branch
  • Fix time_base warnings displayed by FFmpegFrameRecorder (issue #75)
  • Add new TemplateMatching sample, thanks to Waldemar Neto
  • Update instructions in the README.md file for manual installation in Android Studio
  • Replace deprecated CvMat and IplImage functionality used in the Demo class with new Indexer API, and provide sample pom.xml file for Maven
  • Make FFmpegFrameGrabber.getFrameRate() return AVStream.avg_frame_rate instead of r_frame_rate (issue #63)
  • Disable DocLint, which prevents the build from succeeding on Java 8 (issue bytedeco/javacpp#5)
  • Add FlyCapture2FrameGrabber with cross-platform support of FlyCapture2 (pull #45)
  • Fix issue that would prevent CanvasFrame from working on Mac OS X with recent versions of the JDK (issue #39 and issue #314)
  • Upgrade RecordActivity sample with a continuous record loop, thanks to Federico Sendra and Juan Manuel Sobral
  • Make FrameGrabber.createDefault() throw an exception on unsupported input, instead of returning a cryptic null (issue #30)
  • Add videoCodec, videoBitrate, audioCodec, and audioBitrate properties to FrameGrabber
  • Work around avcodec and avdevice not loading properly for FFmpegFrameGrabber and FFmpegFrameRecorder (issue #24)
  • Do key frame detection in FFmpegFrameRecorder based on AVPacket, not AVPicture (pull #20)

July 27, 2014 version 0.9

  • Remove platform property from pom.xml, replaced with the platform.dependency one in JavaCPP Presets (issue #10)
  • Add new RLSA sample, thanks to Nicholas Woodward (issue #469)
  • Fix a timestamp rounding issue in FFmpegFrameGrabber that causes setFrameNumber() to sometimes pick the wrong frame if FPS is not a proper divisor of 1000000 (issue #5)
  • Increase the flexibility of the pom.xml file by making it possible to specify a custom version of JavaCPP
  • Add missing dependencies for JogAmp in the pom.xml file (issue #2)
  • Add new OpenCVFaceRecognizer sample, thanks to Petter Christian Bjelland
  • Add new OpticalFlowDense sample, thanks to Dawit Gebreyohannes (issue #468)
  • Make it easier to try out the FaceRecognition.java sample (issue #1)

April 28, 2014 version 0.8

  • Move from Google Code to GitHub as main source code repository
  • Upgrade support to OpenCV 2.4.9
  • Upgrade supported FFmpeg API to the 2.2 release branch
  • Fix FFmpegFrameRecorder not refreshing the resampler when the format of samples changes (issue #465)
  • Rename the com.googlecode.javacv.cpp package to org.bytedeco.javacpp, and com.googlecode.javacv to org.bytedeco.javacv
  • Removed old NetBeans project files that cause a conflict when trying to open as a Maven project (issue #210)
  • Adjusted the samples a bit because of small changes in the API with the move to the JavaCPP Presets
  • Fixed ObjectFinder not working with recent versions of OpenCV, especially on Android (issue #214)
  • Added new FrameRecorder.gopSize property to let users set a desired GOP size instead of the default one of 12
  • FFmpegFrameGrabber now takes into account calls to setPixelFormat() (issue #429), but does not enforce it
  • Added a Frame.audioChannels field for resampling purposes in FFmpegFrameRecorder (issue #388)
  • In FFmpegFrameRecorder, fixed audio encoding with the Vorbis codec (issue #428) and the WebM container (issue #435), and other audio related things
  • Added missing allocateArray() constructors to CameraParams and MatchesInfo (issue #421)
  • Fixed errors such as "jniopencv_nonfree.dll: Can't find dependent libraries" by adding the opencv_ocl module as dependency
  • Added support to seek in audio-only streams with FFmpegFrameGrabber.setTimestamp() (issue #417)
  • Fixed potential thread concurrency issues and crash in the stopRecording() and onDestroy() methods of the RecordActivity sample, thanks to Jacob Duron
  • To capture the last frame of a video file, reverted FFmpegFrameGrabber.setTimestamp() to its previous behavior (issue #413)
  • Updated samples/FaceApplet.jnlp to make it work with JDK/JRE 7u45

January 6, 2014 version 0.7

  • Upgraded support to OpenCV 2.4.8
  • Upgraded supported FFmpeg API to the 2.1 release branch
  • Updated freenect to reflect the latest changes of OpenKinect's master branch
  • Updated videoInput to reflect the latest changes in the "update2013" branch
  • Added Frame.opaque field to give access to the raw AVFrame in the case of FFmpegFrameGrabber (issue #399)
  • Added new FFmpegFrameGrabber.grabKeyFrame() method to grab key frames (I-frames) directly (issue #312)
  • VideoInputFrameGrabber now uses 640x480 as default image size to prevent "videoInput.getPixels() Error: Could not get pixels."
  • Fixed FFmpegFrameGrabber.setTimestamp() not working for streams with audio (issue #398)
  • Fixed wrong haarcascade_frontalface_alt.xml file getting downloaded by the Demo class (issue #402)
  • Added a Frame.sampleRate field to allow audio samples to be resampled by FFmpegFrameRecorder (issue #388)
  • Incorporated IPCameraFrameGrabber from Greg Perry (issue #384)
  • Fixed thread safety issues with FFmpeg in FFmpegFrameGrabber and FFmpegFrameRecorder (issue #377)
  • Fixed memory leak in the MotionDetector.java sample file (issue #372)
  • New videoCodecName and audioCodecName properties to allow users of FFmpegFrameRecorder to use codecs such as "libx264rgb" (issue #369)

September 15, 2013 version 0.6

  • Upgraded supported FFmpeg API to the 2.0 release branch (with Java interface files now based on code automatically produced by JavaCPP Presets)
  • Fixed FFmpegFrameGrabber.getFrameNumber()
  • Upgraded support to OpenCV 2.4.6
  • Fixed callbacks when used with custom class loaders such as with Web containers
  • Upgraded to ARToolKitPlus 2.3.0 (issue #234)
  • Fixed drawing issues with MarkerDetector.draw()
  • Fixed FFmpegFrameGrabber.getTimestamp() not returning values for audio frames (issue #328)
  • Added new Frame.keyFrame field returned by FFmpegFrameGrabber.grabFrame() to know when a grabbed frame is a key frame or not (issue #312)
  • Worked around problem in samples/RecordActivity.java that would happen when trying to record a frame with an invalid timestamp (issue #313)
  • Fixed potential resource leak that could occur after FFmpegFrameRecorder throwing an Exception
  • Fixed FFmpegFrameGrabber not returning the last few frames of video streams (issue #315)
  • Fixed wrong dependencies of OpenCV preventing correct loading (issue #304)
  • Renamed FrameRecorder.record(Buffer[] samples) to a cleaner record(Buffer ... samples) (issue #303)
  • Fixed FFmpegFrameRecorder not flushing buffers on stop() (issue #302)

April 7, 2013 version 0.5

  • Upgraded support to OpenCV 2.4.5
  • Upgraded supported FFmpeg API to the 1.2 release branch
  • New methods FFmpegFrameRecorder.setVideoOption() and setAudioOption() generalize the way to set arbitrary codec options, such as "profile", "preset", "tune", etc. used by the x264 codec
  • Included better format guessing inside FFmpegFrameRecorder for protocols like RTP
  • Added support for planar audio formats to FFmpegFrameGrabber and FFmpegFrameRecorder, as required by newer versions of FFmpeg for at least MP3 and AAC
  • Enhanced FFmpegFrameRecorder by making it use the closest supported frame rate for the given codec instead of failing
  • To support variable bitrate (VBR) encoding, appended new videoQuality and audioQuality properties to FFmpegFrameRecorder, which usually have an effective range of [0, 51] and overrides the videoBitrate and audioBitrate properties

March 3, 2013 version 0.4

  • Upgraded support to OpenCV 2.4.4
  • CanvasFrame.waitKey(-1) does not wait anymore and returns the last KeyEvent dispatched since the last call to it
  • Upgraded supported FFmpeg API to the 1.1 release branch
  • Fixed bug in FaceRecognition.java sample (issue #276)
  • Included Sobel(), Scharr(), Laplacian(), and Canny() from opencv_imgproc whose equivalent functions in the C API have missing parameters
  • Extended OpenKinectFrameGrabber with setDepthFormat() and setVideoFormat() methods to be able to set both formats independently (issue #273)
  • Fixed Blender.blend() having its @OutputMat incorrectly annotated as @InputMat (issue #272)
  • Added new RecordActivity.java Android sample from Shawn Van Every and Qianliang Zhang
  • Added missing allocate() methods for FunctionPointer in AVIOContext and others, which prevented these FFmpeg callbacks from functioning
  • Fixed infinite loop in FrameGrabber.Array.grab() (as used by ProCamCalib in the case of stereo cameras, issue #262) when FrameGrabber.getTimestamp() returns an invalid negative value (as with opencv_highgui) or when using different types of (unsynchronized) FrameGrabber together
  • Fixed cvQueryHistValue_1D() and other functions that use a raw CvArr object
  • Fixed problem when subclassing CanvasFrame

November 4, 2012 version 0.3

  • Upgraded support to OpenCV 2.4.3 (issue #233)
  • Fixed functions like Algorithm.getMat() and HOGDescriptor.getDefaultPeopleDetector() returning null instead of the expected data
  • Implemented better, more transparent, handling of cv::Ptr
  • When allocating an empty IplImage, CvMat, CvBGCodeBookModel, etc. its memory content now gets zeroed out, giving OpenCV a better chance of displaying an error message instead of crashing
  • Upgraded supported FFmpeg API to the 1.0 release branch
  • Appended to StringVector and MatVector new convenient bulk constructors and put() methods taking arrays of String, IplImage, CvMat, etc.
  • Included new Blobs module from David Grossman and the corresponding BlobDemo sample
  • Added missing opencv_core.partition() function (issue #144)
  • Fixed up the samples a bit (issue #229 and issue #230)
  • Switched the majority of @Adapter annotations to more concise ones like @StdVector as allowed by new capabilities of JavaCPP
  • Fixed FFmpegFrameGrabber.getLengthInFrames() and OpenCVFrameGrabber.getLengthInTime() (issue #231 and issue #236)
  • Enhanced FFmpegFrameRecorder to support conversion between audio sample formats (for the experimental AAC encoder among other things) and to let two different threads call record(samples) and record(image) simultaneously, plus a couple of other features like setFrameNumber(), which lets users skip image frames (achieving variable frame rate)
  • Added a javacpp.skip property to pom.xml, such that a command like mvn package -Pall -Djavacpp.skip=true only recompiles the Java source files, but also added platform.root and compiler.path properties, which map directly to JavaCPP's for convenience

July 21, 2012 version 0.2

  • Provided new javacv-linux-arm.jar build thanks to Jeremy Nicola (issue #184)
  • Additional default properties inside pom.xml make it easier to build JavaCV from source (issue #202), calling mvn package now succeeds with only OpenCV and a C++ compiler for JavaCPP
  • Made a few minor updates for OpenCV 2.4.2
  • New Pointer.limit property of JavaCPP can now be used to get the size of an output parameter, and to specify the maximum size on input as well
  • Upgraded supported FFmpeg API to the 0.11 release branch
  • Added audio support to FFmpegFrameGrabber (call grabFrame() instead of grab()) and FFmpegFrameRecorder (call setAudioChannels() before start(), and record(Frame) instead of record(IplImage)) (issue #160)
  • Gave better default FFmpegFrameRecorder settings to H.263, MPEG-4, etc. codecs and fixed H.264 encoding with libx264 (issue #160)
  • Refined the FaceApplet sample
  • Fixed FlannBasedMatcher constructor, FaceRecognizer.train(), and Stitcher.stitch()/composePanorama() (issue #211)
  • Fixed CanvasFrame sometimes blanking out under Windows and maybe Linux (issue #212)

May 27, 2012 version 0.1

  • Started using version numbers, friendly to tools like Maven, and placing packages in a sort of Maven repository
  • JavaCV can now extract and load native dependent libraries such as libopencv_core.so.2.4, libopencv_core.2.4.dylib, opencv_core240.dll, etc. from Java resources placed inside the com.googlecode.javacv.cpp.<platform.name> package (i.e.: under the /com/googlecode/javacv/cpp/<platform.name>/ directory of a JAR file in the classpath) (issue #146)
  • Included new FaceApplet sample to demonstrate How to use JavaCV in an applet
  • Added handy IplImage.asCvMat() and CvMat.asIplImage() conversion methods
  • Fixed a few small things with OpenCVFrameGrabber, opencv_contrib, opencv_legacy, and opencv_stitching

May 12, 2012

  • Upgraded support to OpenCV 2.4.0 (issue #187)
  • Moved the source code repository to Git
  • Added pom.xml file for Maven support and changed the directory structure of the source code to match Maven's standard directory layout
  • Made it easier to create one massive statically linked native library by passing something like "-Xcompiler -Wl,-static -o javacv" as command line options to JavaCPP, usually from inside build.xml or pom.xml (issue #146)
  • Fixed missing parameter from CvANN_MLP.create()
  • Added methods cvCalcCovarMatrixEx(), cvEigenDecomposite(), and cvEigenProjection() taking an IplImage[] as argument for convenience
  • VideoInputFrameGrabber.start() now accepts a connection argument such as VI_COMPOSITE to support analog cameras and what not
  • Fixed FaceRecognition sample (issue #188)
  • Added a few convenience methods to avoid the need to create empty CvAttrList

March 29, 2012

  • Added missing array allocators and position() methods to KDTree.Node, DefaultRngAuto, CvAffinePose, KeyPoint, BaseKeypoint, ReferenceTrees, DMatch, *.Params, CvFuzzy*, Octree.Node, CvDefParam, Cv*Blob*, Cv*Track*, CvDrawShape, CvVectors, CvParamGrid, Cv*Params, CvSVM*, CvPair16u32s, CvDTree* CvTrainTestSplit, CvMLData, FeatureEvaluator, and *DataMatrixCode
  • Increased versatility of IplImage.createFrom(), copyFrom(), copyTo(), getBufferedImage() by providing a flipChannels parameter, whose effect was previously mistakenly forced onto four-channel images of byte values only (issue #163)
  • Fixed a couple of things with CvMat.get()/put() (issue #167)
  • In addition to an IplImage, we may now specify the pixel format of the data when calling FFmpegFrameRecorder.record(), but otherwise when IplImage.nChannels == 2, it assumes PIX_FMT_NV21, allowing for easy and efficient encoding of data captured from the camera on Android (issue #160), image objects we can also convert to RGB using cvCvtColor() with CV_YUV420sp2BGR
  • Fixed seeking capabilities of FFmpegFrameGrabber (issue #162) and added getLengthInFrames() and getLengthInTime() methods to query the duration of streams, when known
  • Enhanced IplImage.clone() and create*Compatible() with cloning of their BufferedImage to make it easier to keep color components in the right order (issue #163)
  • Refactored FrameGrabber and FrameRecorder a bit to accommodate new createDefault(...) and create(String className, ...) factory methods, offering to users an easier selection method to work around limitations of some APIs (issue #70)
  • Adjusted GNImageAligner, ProCamTransformer, etc. to support alignment of only the projector display on textureless surface planes
  • Renamed a few more Settings properties to reflect better their meanings

February 18, 2012

  • Added GLCanvasFrame to show OpenGL renderbuffers on screen, plus a new factory method JavaCVCL.createCLGLImageFrom() to create compatible ones from IplImage objects, as well as more user-friendly getGLContext(), getGL() and getGL2() methods
  • Fixed various things of the original CanvasFrame, and JavaCV.createCLImage() and createIplImage(), also appending From to their names
  • New createPinnedBuffer() and createPinnedIplImage() factory methods in JavaCVCL to allocate page-locked memory for faster CPU<->GPU transfers, but it does not seem to work for OpenCL image objects, only linear buffer objects :(
  • Fixed and enhanced GNImageAlignerCL and ProjectorDevice (its useOpenGL property) to support fully OpenCL and OpenGL acceleration
  • Refactored Parallel a bit so that we may set the number of threads it uses via its static numThreads property or the "com.googlecode.javacv.numthreads" system property, which defaults to Parallel.getNumCores() = Runtime.getRuntime().availableProcessors()
  • Cleaned up and renamed some methods in JavaCV, while adding boundingRect(), functionally similar to cvBoundingRect, but better adapted to compute a properly aligned and padded ROI
  • Inserted a couple of missing allocate() inside opencv_flann
  • Updated ObjectFinder with a Settings.useFLANN property to let it use FLANN via OpenCV
  • Cleaned up and optimized HandMouse
  • CanvasFrame, FrameGrabber, FrameRecorder, and ProjectiveDevice objects now throw Exception objects of a nested class instead of the too generic java.lang.Exception one
  • Moved parallel execution of cvkernels.multiWarpColorTransform(), modifying ImageTransformer classes, from GNImageAligner into cvkernels, which now also supports other image types than float
  • Renamed some Settings properties here and there to correct typos and reflect better their meanings
  • Updated freenect to reflect the latest changes of OpenKinect's master branch
  • FFmpeg and other libraries did not work under Android when compiled with the latest NDK, r7 (issue #147): Fixed in JavaCPP
  • Moved IplImage.timestamp to FrameGrabber, also adding a frameNumber property, both allowing to seek within streams too
  • Removed triggerFlushSize property from CameraDevice and FrameGrabber, instead relying on the numBuffers property to decide the required size of a buffer flush
  • Corrected the logic behind FFmpegFrameGrabber.getFrameRate() and getTimestamp() (issue #151)
  • Created a BufferRing class for convenient circular rings of large buffers that require manual release of resources, such as OpenCL memory
  • Added a few more useful methods to FrameGrabber, including restart(), flush(), and delayedGrab() (to be used in conjunction with getDelayedTime() and getDelayedImage())
  • Inserted cvLoadImageBGRA() and cvLoadImageRGBA() methods into opencv_highgui to load color images compatible with OpenCL more easily
  • JavaCvErrorCallback now outputs messages to Logger instead of System.err
  • Defined VI_COM_MULTI_THREADED for videoInput, allowing it to run on multiple threads if needed

January 8, 2012

  • JavaCV should now have an easier time automatically finding libraries inside standard directories such as /usr/local/lib/, /opt/local/lib/, and C:\opencv\, even when they are not part of the system configuration or PATH (issue #127)
  • Renamed set() and fill() methods to put() inside CvPoint* classes, for better naming consistency
  • Renamed FrameGrabber.ColorMode to ImageMode and its BGR value to COLOR to reflect the fact that a FrameGrabber instance can return color images in some arbitrary format, but added a new pixelFormat property to let users know or specify the exact pixel format desired, such as PIX_FMT_BGR24, etc. in the case of FFmpegFrameGrabber
  • After FFmpegFrameGrabber.start(), the format, imageWidth, imageHeight, and frameRate properties switch to their effective values
  • Added new FrameGrabber.sensorPattern property to obtain the Bayer filter layout of raw data from DC1394FrameGrabber and FlyCaptureFrameGrabber
  • Readded to KDTree, Index, and HOGDescriptor some functions with FloatPointer and IntPointer arguments that were mistakenly removed when OpenCV switched to using cv::InputArray and cv::OutputArray parameter types (issue #134)
  • Renamed ProjectiveGainBiasTransformer to ProjectiveColorTransformer
  • Added a few classes to do some processing using OpenCL and OpenGL: JavaCVCL, GNImageAlignerCL, ProjectiveTransformerCL, ProjectiveColorTransformerCL, and ProCamTransformerCL with some other related files
  • Renamed Parallel.numCores to the more conventional Parallel.NUM_CORES
  • Added new FaceRecognition.java sample from Stephen L. Reed
  • Inserted a couple of missing calls to Loader.load() (issue #142)
  • Improved hacks for Loader.load() in JavaCPP make JavaCV work on Android 4.0
  • New PS3EyeFrameGrabber from Jiri Masa can now grab images using the SDK from Code Laboratories

October 1, 2011

  • Fixed DC1394FrameGrabber and FlyCaptureFrameGrabber to behave as expected with all Bayer/Raw/Mono/RGB/YUV cameras modes (within the limits of libdc1394 and PGR FlyCapture) (issue #91)
  • Fixed regression of IplImage.copyFrom() and createFrom() with BufferedImage objects of SinglePixelPackedSampleModel (issue #102)
  • C++ functions using std::vector objects as output parameters now work on Windows Vista and Windows 7 as well

August 20, 2011

  • Upgraded support to OpenCV 2.3.1
  • An output argument of type cv::Mat or cv::OutputArray returned with a size 0 now correctly sets CvArr.address = 0
  • Fixed IplImage.createFrom() and copyFrom() when called on objects returned by BufferedImage.getSubimage()
  • Added missing allocator to CvRNG
  • OpenCVFrameGrabber now detects when CV_CAP_PROP_POS_MSEC is broken and gives up calling cvGetCaptureProperty()
  • New OpenKinectFrameGrabber.grabDepth() and grabVideo() methods to capture "depth" and "video" simultaneously, regardless of the mode

July 5, 2011

  • Upgraded support to OpenCV 2.3.0
  • Fixed OpenKinectFrameGrabber, which can now also capture depth images when setFormat("depth") is called before start()
  • Fixed CvMatArray and IplImageArray as well as histogram related functions
  • Fixed FFmpegFrameGrabber, and FFmpegFrameRecorder now works on Android also
  • Fixed calls, such as opencv_flann.Index.knnSearch(), that require a MatAdapter or an ArrayAdapter for output

June 10, 2011

  • New freenect wrapper and corresponding OpenKinectFrameGrabber to capture from Microsoft's Kinect stereo camera using OpenKinect
  • JavaCV now exposes all C++ functions and classes of OpenCV not covered by the C API
  • Fixed various erroneous declarations and calls, including those due to changes in JavaCPP

May 11, 2011

  • Removed CvMat object pooling in favor of more efficient ThreadLocal objects created by CvMat.createThreadLocal()
  • Changed Marker.getCenter() back to the centroid, because it has better noise averaging properties and gives in practice more accurate results than the actual center
  • Added hack to OpenCVFrameGrabber.start() to wait for cvRetrieveFrame() to return something else than null under Mac OS X
  • FFmpeg now works properly on Windows and Android (issue #63) with newer binaries
  • New videoInputLib wrapper and corresponding VideoInputFrameGrabber to capture using DirectShow, useful under Windows 7 where OpenCV and FFmpeg can fail to capture using Video for Windows (issue #58)
  • GeometricCalibrator now reports the maximum errors in addition to the average (RMS) errors

April 7, 2011

  • Added a format property to CameraDevice, FrameGrabber, and FrameRecorder, mostly useful for FFmpegFrameGrabber, where interesting values include "dv1394", "mjpeg", "video4linux2", "vfwcap", and "x11grab"
  • OpenCVFrameRecorder now uses CV_FOURCC_PROMPT under Windows as default since CV_FOURCC_DEFAULT crashes (issue #49)
  • Added hack to make sure the temporarily extracted library files get properly deleted under Windows
  • JavaCPP now loads classes more lazily
  • Fixed most occurences of UnsatisfiedLinkError (issue #54), but some corner cases may require a call to Loader.load() on the class one wishes to use
  • Added (rudimentary) outlier detection and modified zero threshold handling in the image alignment framework
  • New JavaCV.hysteresisThreshold() feature
  • New HandMouse functionality, which depends on the image alignment framework
  • Fixed ProjectiveDevice.distort(), which mistakenly undistorted images instead
  • New HoughLines sample thanks to Jeremy Nicola

February 19, 2011

  • Switched from JNA to JavaCPP, which has a lower overhead and supports C++, bringing hope that future versions of JavaCV will support features of OpenCV available only through the C++ API
  • Consequently, the syntax of various operations have changed a bit, but the transition should not be too painful
  • As a happier consequence, this also fixes the problem with SSE instructions on 32-bit x86 (issue #36)
  • Also, JavaCPP does not have any limitations or performance issues with large data structures (issue #10 and issue #14)
  • Added support for OpenCV 2.2 (issue #42), but dropped support for all previous versions
  • Added samples provided by users (issue #1, issue #45, and issue #46)
  • Added deinterlace setting to FFmpegFrameGrabber having it call avpicture_deinterlace() (issue #38)
  • Enhanced a few things of the image alignment algorithm
  • Tried to fix image format conversion inside FlyCaptureFrameGrabber, but this is going to require more careful debugging
  • Fixed and added various other things I forget

December 2, 2010

  • Now works on Android with the Dalvik VM (for more details, please refer to the FacePreview sample available on the download page)
  • Added more hacks to CanvasFrame in the hope to make it behave better outside the EDT
  • Made clearer the error messages thrown from FrameGrabber objects, when start() may not have been called
  • Fixed version specific declarations of CvStereoBMState and related functions
  • Fixed conditions that could crash cvkernels

November 4, 2010

  • Renamed the package namespace to com.googlecode.javacv, which makes more sense now that JavaCV has been well anchored at Google Code for more than a year, piggybacking on the unique and easy-to-remember domain name
  • Included new FFmpeg wrapper classes avutil, avcodec, avformat, avdevice, avfilter, postprocess, and swscale, eliminating the need of the separate FFmpeg-Java package
  • CanvasFrame now redraws its Canvas after the user resizes the Frame
  • Fixed the Error thrown when calling CanvasFrame.showImage() from the EDT
  • Added check to DC1394FrameGrabber so that a "Failed to initialize libdc1394" does not crash the JVM
  • FFmpegFrameGrabber does not crash anymore when forgetting to call start() before a grab() or trigger()
  • FrameGrabber now selects the default grabber a bit better
  • Made sweeping changes (for the better, but still not finalized) to GNImageAligner, ProjectiveTransformer, ProjectiveGainBiasTransformer, and ProCamTransformer...
  • Added to JavaCV more methods related to transformation of planes: perspectiveTransform(), getPlaneParameters(), getPerspectiveTransform(), and HtoRt(), as well as ProjectiveDevice.getFrontoParallelH()
  • Added a static autoSynch flag to all Structure classes of cxcore, cv, and cvaux, which you may set to false prior to the return of things like big and heavy CvSeq to make them load faster and to avoid stack overflows, but accessing fields will then require manual calls to readField() and writeField() (issue #10 and #14)
  • Added missing ByValue subclasses to CvSeq, CvSet, CvContourTree, and CvChain... Any others missing?
  • Fixed Exception thrown from cvCreateHist() under JNA 3.2.7 (issue #26)
  • Enhanced CvMat.put(), which now supports setting submatrices
  • Improved inside IplImage the support of BufferedImage, especially those using a DirectColorModel (issue #23)
  • Fixed crash in cvkernels when color transformation X is null

July 30, 2010

  • Fixed crash that would occur in CanvasFrame for some video drivers
  • FFmpegFrameGrabber now supports other input formats (devices), such as x11grab that can be used for screencasting
  • Added JavaCV.median() function, and JavaCV.fractalTriangleWave() now respects image ROI
  • Fixed background subtraction in cvaux
  • Fixed crash inside the code for direct alignment caused by the ROI getting set outside the image plane
  • Added deltaScale and tryToFixPlane to GNImageAligner.Settings (the first used in ImageTransformer.Parameters as increment, randomly selected forward or backward, for finite difference), which sometimes help to jump over local minima

May 30, 2010

  • Removed redundant CvMemStorage.clearMem() method, use cvClearMemStorage()
  • Fixed the sample Test2 class that did not work under Windows
  • Fixed corruption by the cvkernels transformer at the borders
  • Modified CanvasFrame constructors and added a gamma argument used by showImage(IplImage)
  • CanvasFrame now lets users resize the frame, while displayed images are stretched to fit the new size
  • Renamed CanvasFrame.acquireGraphics() to createGraphics() for consistency
  • When FlyCaptureFrameGrabber cannot set fastest speed, it now safely fails by setting any supported speed
  • Added a new Parallel.loop() method that can use more threads than the number of CPU cores detected
  • Added new numThreads property to GNImageAligner and fixed a few minor inconsistencies as well
  • Fixed incorrect Java.HnToRt(), and added a few norm() and randn() methods
  • For functions with float[] and double[] arguments in cvaux and cv, added complementary FloatBuffer and DoubleBuffer declarations
  • Fixed loading problems with cvaux
  • Fixed and enhanced histogram, back projection, and other CAMSHIFT related functionality
  • Added code for CvRNG
  • Added "/opt/local/lib/" and "/opt/local/lib64/" (standard on Mac OS X) to the default list of search paths for OpenCV
  • Added CvScalar.getVal() and CvIntScalar.getVal(), which simply return the val field, convenient for Scala where val is a reserved word
  • Fixed the construction of IplImage from a Pointer
  • Removed incorrect cases when an IplImage gets converted to a BufferedImage.TYPE_CUSTOM
  • Made CvArr.PointerByReference a bit more consistent and general

April 16, 2010

  • Modified IplImage, FrameGrabber, and CanvasFrame to get better default behavior of gamma correction
  • Fixed cv.CvHistogram and related histogram functions
  • CameraDevice.Settings.triggerFlushSize now defaults to 5 (only affects OpenCVFrameGrabber and FFmpegFrameGrabber)
  • Replaced LMImageAligner by GNImageAligner, a more appropriate name for Gauss-Newton with lineSearch
  • Fixed a few things related with ProjectiveDevice.Settings

April 8, 2010

  • Added support for OpenCV 2.1

April 5, 2010

  • Fixed up clone() methods to avoid the need to cast
  • Removed the fullScreen argument from CanvasFrame constructors, which will now switch to full-screen mode only when a screenNumber is explicitly passed
  • Renamed FrameGrabber.ColorMode.GRAYSCALE to GRAY
  • Replaced deprecated functions from FFmpegFrameGrabber and FFmpegFrameRecorder
  • FFmpegFrameGrabber can now resize images

March 21, 2010

  • Added new classes and methods used by ProCamTracker: cvkernels, JavaCV.fractalTriangleWave(), ImageAligner, LMImageAligner, ImageTransformer, ProjectiveTransformer, ProjectiveGainBiasTransformer, ProCamTransformer, and ReflectanceInitializer
  • CameraDevice.Settings has a new deviceFile property (used by a FrameGrabber), which brings up a file dialog for some PropertyEditors
  • Moved in CameraSettings, ProjectorSettings, and FrameGrabber.PropertyEditor from the procamcalib package
  • Added to CameraDevice.Settings and FrameGrabber a triggerFlushSize property to indicate the number of buffers to flush on trigger() to compensate for cheap cameras that keep old images in memory indefinitely
  • Changed the type of CameraDevice.Settings.deviceNumber to Integer so we may set it to null
  • Fixed and enhanced CanvasFrame.showImage() methods a bit
  • In triggerMode DC1394FrameGrabber now tries to use a real software trigger and only falls back to one-shot mode on error
  • Fixed array constructors of IplImage.PointerByReference() and CvImgObsInfo.PointerByReference()
  • Added CvPoint.fillArray() methods to reuse preallocated arrays and changed createArray() a bit as well
  • Fixed and enhanced all IplImage.copy*() methods, including new support for ROIs and subimages, which affects create*() and getBufferedImage() methods as well
  • Updated Marker to support different size and spacing in X and Y
  • Added Settings to ObjectFinder
  • Fixed distortion problem in ProjectiveDevice and ProCamColorCalibrator with OpenCV 1.1pre1
  • Split ProjectiveDevice.Settings into ProjectiveDevice.CalibrationSettings (for applications like ProCamCalib) and ProjectiveDevice.CalibratedSettings (for applications like ProCamTracker)
  • Renamed gamma to responseGamma in ProjectiveDevice, and moved previous nominalDistance parameter to Settings
  • Added ProjectiveDevice.rescale() to rescale calibration parameters when switching a device to a new image size
  • ProjectiveDevice.undistort() and distort() can now useFixedPointMaps of OpenCV
  • ProjectiveDevice and its subclasses now throw new Exception() if the parameterFile cannot be read

February 13, 2010

  • Relicensed JavaCV under the GPLv2 with Classpath exception (see LICENSE.txt). Please note that if your application links with code that needs ARToolKitPlus, for example, it will become subject to the full GPL, without Classpath exception
  • Added devicePath setting to CameraDevice that works with FFmpegFrameGrabber, OpenCVFrameGrabber, and other FrameGrabber with a String constructor
  • Added "C:/OpenCV2.0/bin/release/" to the directory list to search for OpenCV DLLs
  • Moved cvFindHomography(), cvFindExtrinsicCameraParams2(), cvReprojectImageTo3D(), cvSaveImage(), and cvRetrieveFrame() to version specific classes since their number of arguments differ with the version of OpenCV
  • Enhanced CvMat.put(CvMat mat) to work better even when the matrices are not actually compatible
  • Added new IplImage factory methods createCompatible(IplImage image), createIfNotCompatible(IplImage image, IplImage template), and createFrom(BufferedImage image)
  • Fixed distortionCoeffs corruption that might occur in ProjectiveDevice

January 3, 2010

  • Added wrapper for the cvaux module of OpenCV
  • Added abstract FrameRecorder class and a OpenCVFrameRecorder class
  • Fixed read() problem that might occur within Pointer constructors
  • Running java -jar javacv.jar now displays version information

December 22, 2009

  • Fixed CanvasFrame from getting stuck in a maximized window
  • Removed all setAutoWrite(false) from cxcore now that the bug appears fixed in JNA
  • Added FFmpegFrameGrabber and FFmpegFrameRecorder to easily record live footage and grab back offline into JavaCV

November 24, 2009

  • Added more convenient constructors and factory methods for CvPoint*, CvSize*, CvRect, CvTermCriteria, CvSlice, and CvAttrList
  • Added R2 correlation coefficient field to ProjectiveDevice
  • Enhanced and fixed color conversion spaghetti code in FlyCaptureFrameGrabber
  • Fixed the CvHaarFeature Structure
  • Renamed CvIntScalar factory methods to match with CvScalar
  • Enhanced and fixed some problems with gamma correction in IplImage
  • Added a highgui.CV_FOURCC() method that takes chars as parameter
  • Moved MarkedPlane.drawMarkers() to Marker.draw() for better code reuse
  • Added MarkedPlane.getTotalWarp() with a "useCenters" parameter
  • Changed default values of MarkerDetector.binarizationKWhiteMarkers to 1.0 and ProjectorDevice.brightnessBackground to 0.0
  • Fixed issue with image width and memory alignment in MarkerDetector
  • Marker.getCenter() now computes the actual physical center instead of the centroid
  • OpenCVFrameGrabber.getDeviceDescriptions() now throws UnsupportedOperationException
  • Added support in OpenCVFrameGrabber to grab frames from video files
  • Added ProjectiveDevice.getRectifyingHomography() method
  • Added JavaCvErrorCallback to easily catch errors of OpenCV in Java

October 19, 2009

  • Moved the functionality of CvMatPool to the CvMat.take() and .pool() methods
  • Added color calibration for projector-camera systems (ProCamColorCalibrator)
  • Updated DC1394FrameGrabber to handle more conversion use cases automatically
  • Fixed CvIntScalar to mirror CvScalar

October 14, 2009

  • Change of plan: JavaCV now works with any of OpenCV 1.0, 1.1pre1, or 2.0! Version specific functionality is enclosed in subclasses, e.g., the class cv.v20 can access everything from the cv module of OpenCV 2.0
  • Added a few missing functions and adjusted some mappings to make them closer to the C API
  • Added a few more helper methods to CvPoint*
  • Added temporary storage to ObjectFinder to plug the memory leak

October 2, 2009

  • Fixed problem when loading distortion coefficients with ProjectiveDevice
  • Added automatic read and write for functions with arrays of Structure or PointerByReference
  • Added to cv.java a few missing functions related to calibration
  • Fixed up a bit helper methods for CvPoint*, CvScalar, CvRect, CvBox2D, CvMat, IplImage, CvMemStorage, CvSeq, and CvSeqBlock
  • Added CvMatPool to MarkedPlane and Marker
  • Added a few new distort() methods to ProjectiveDevice
  • Last version to support OpenCV 1.1pre1: Future version will require OpenCV 2.0

August 27, 2009

  • IplImage now flips the buffer on copy if necessary
  • Added needed Pointer constructor for CvSURFPoint and CvConvexityDefect
  • Cleaned up a bit the messy Buffers in CvMat

August 26, 2009

  • Added get*Buffer() functions to IplImage
  • Added more options for gamma correction in IplImage and ProjectiveDevice
  • Further cleaned up the namespace and constructors of ProjectiveDevices
  • CanvasFrame.waitKey() now only checks KeyEvent.KEY_PRESSED
  • Added CvMatPool to avoid recreating matrices
  • Moved CvScalar functions to cxcore

August 19, 2009

  • Switched to using import static for relief from namespace hell
  • Fixed color channel reversal of Bayer images in DC1394FrameGrabber

August 11, 2009

Initial release

Acknowledgments

This project was conceived at the Okutomi & Tanaka Laboratory, Tokyo Institute of Technology, where I was supported for my doctoral research program by a generous scholarship from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of the Japanese Government. I extend my gratitude further to all who have reported bugs, donated code, or made suggestions for improvements (details above)!