OpenCV 4.12.0

What's new in this version:

Core Module:
- Added user-defined logger callback
- Added reinterpret() method to cv::Mat
- Fixed empty ND-array construction
- Fixed bug with int64 support for FileStorage
- Fixed cv::meanStdDev overflow for large images
- Useed HAL for cv::mean function too
- Vectorized cv::normalize / cv::norm
- Enabled SIMD_SCALABLE for exp and sqrt
- vectorized normDiff with universal intrinsics
- Refactored normDiff in hal_rvv and extended with support of more data types
- Further vectorized copyTo with mask
- Fix POWN OpenCL implementation
- Restored legacy universal intrinsics operators in a separate header
- Improved solveCubuc accuracy:
- Deprecated copyData Parameter in cv::UMat construction from std::vector and always copy data

Imgproc Module:
- Optimized and reduced memory consumption in cv::findContours
- Fixed integer overflow in cv::HoughLinesProbabilistic with very large input images
- Fixed several potential overflows in image warping
- Improved robustness for ellipse fitting
- Added cv::THRESH_DRYRUN flag to get adaptive threshold values without input thresholding
- Fixed cv::getPerspectiveTransform for singular case
- Optimized bayer2Gray with universal intrinsics
- Added optional mask to cv::threshold
- Accelerated cv::remap calculation with multi-threading, when cv::WARP_INVERSE_MAP is used
- Reworked HSV color conversion tables initialization in OpenCL branch to fix possible data raise
- Fixed bilateral filter's sigma color and sigma space issue
- Impoved performance of medianblur
- Added cv::getClosestEllipsePoints function to get the closest point on an ellipse
- Corrected IPP implementation for cv::distanceTransform results with single thread
- Fixed cv::matchTemplate with mask crash
- Added diamond structuring element to morphology operations
- Improved cv::intersectConvexConvex accuracy in corner cases

Calib3d module:
- Fixed inliers updating in homography estimation
- Added cv::solvePnPRansac implementation for the fisheye camera model
- Fixed different marker ordering with cv::findChessboardCornersSBWithMeta and cv::CALIB_CB_LARGER flag.
- Fixed logical mistake in FilterSpecklesImpl
- Optimized undistort points for the fisheye camera model
- Optimized new camera matrix estimation and some related functions
- Added warning if projected axes are out of camera frame in cv::drawAxes function
- Added image dimension check to avoid StereoSGBM non-determinism
- Fixed bugs in SQPNP orthogonalization and simplified column vectors copying

DNN Module:
- Added TFLite StridedSllice (without strides but just Slice) and enable strides for TF importers
- Added fusion handler of conv+eltwise in case of multi-output node (i.e. Split)
- Added new ops import in TFLite parser: SUB, SQRT, DIV, NEG, SQUARED_DIFFERENCE, SUM
- Fxied scalars and 1d tensors support in TFLite parser
- Reduced mumber of NHWC<->NCHW conversions in TFLite parser
- Added checks before getting latest pin in Net::Impl::getLatestLayerPin
- Added OpenVINO NPU support
- Added more operations to CANN backend

Objdetect module:
- Extend ArUcoDetector to run multiple dictionaries in an efficient manner
- Fixed rotated aruco marker board generation
- Fixed incorrect Aruco marker detection near image edge
- Fixed invalid vector access in QR de/encoder
- Fix QR code encoder with auto version
- Aded QRCode ECI encoding support
- Made ChAruco board consistency check optional

Photo Module:
- Added 16-bit support to cv::fastNlMeansDenoising and updated tests
- Local decolor pipeline optimization
- Fixed Triangle Weights used by Debevec algorithm being non zero at extremes
- Fixed NaNs in HDR Triangle Weights and Tonemapping and Update LDR Ground Truth in tutorial

VideoIO Module:
- Fixed invalid custom FFmpeg options handling provided with environment variable
- Fixed default cv::CAP_PROP_ORIENTATION_ATUO behaviour
- Improved camera opening performance when with DShow backed, if custom options are provided
- Added zoom support into Android native camera capture
- Fixed V4L frame size for non-BGR output
- Added Orbbec Gemini 330 camera support
- Fixed cv::VideoWriter failure on writing colorless images
- Fixed Android setCameraIndex issue

Imgcodecs:
- Extended image I/O API with metadata support
- Added in-memory animation encoding and decoding
- Added hidden frame support for Animated PNG
- Added GIF decode and encode for imgcodecs
- Added animated WebP Support
- Added GDAL multi-channel support
- Fixed AVIF linkage on Windows
- Significantly improved PNG and Animated PNG files handing
- Added several fixes for libspng integration
- Fixed memory leak in JpegXL decoder
- Added support for cv::IMREAD_UNCHANGED and other ImreadFlags into JpegXL implementation
- Implemented imdecode() for JpegXL to directly read from memory
- Added lossless compression support for JpegXL
- Fixed cv::imread() implementation for 16-bit png with libspng back-end

Highgui Module:
- displayOverlay doesn't disappear after timeout
- Fix closing of windows when using the Qt backend
- Fixed QT window resize regression

G-API module:
- Workaround for successfull append of OpenVINO Execution Provider: Moved creation of 'Ort::Env' before it
- Adding AddRgbFeature(), and improving robustness in ComputeRgbDistance
- Added MS Media Foundation availability check in G-API too

Video Module:
- Added trackers factory with pre-loaded dnn models

HAL:
- Extracted HAL implementations as dedicated folder
- Implemented new HAL for RISC-V RVV 1.0 platforms, see all changes by query
- Refactored OpenVX implementation as HAL
- Partially refactored Intel IPP integration as HAL
- Fixed HALs dependency from thre cv namespace
- Updated KleidiCV HAL for ARM to version 0.5
- Extended FastCV HAL for Qualcomm platforms
- Added new HAL entries for cv::sum

Python, Java and JavaScript bindings:
- Added conditional inclusion support to header parser
- Updated predefined_types.py
- Added bindings for animations
- Fix wrong python type hints for cv::imread
- Replaced tostring() and fromstring() call in tests and samples for compatibility with modern Python
- Replaced ndarray.ptp() with np.ptp() for NumPy 2.0 Compatibility
- Extended Java bindings generation in DNN
- Fixed std::vector<Point3f> handling in JS wrappers
- Added cv::VideoCapture buffered stream constructor into Java
- Extended bindings for cv::HoughLinesWithAccumulator
- Fixed Kotlin 2.0 internal error for unsafe coercions

Platforms support:
- Fixed OpenBLAS detection on Linux
- Fixed compatibility with CMake 4
- Fixed Android build with Vulkan support
- Fixed VSX intrinsics (Power) implementation
- Added CMake checks for getauxval and elf_aux_info for POWER architecture
- Fixed build for openBSD
- Added initial support for Nvidia Blackwell GPU arch in CUDA
- Fixed cv namespace dependency in Carotene HAL
- Fixed _DEBUG/NDEBUG handling across modules
- Fixed build for QNX 7.0
- Switched to statically linked FastCV
- Forced C++17 Standard for CUDA targets when CUDA Toolkit >=12.8
- Fixed build for Windows ARM64EC
- Fixed filesystem support under Cygwin environment