OpenCV

What's new in this version:

Generic:
- Internal C API cleanup and back-ports for 5.x
- RISC-V/AArch64: disable CPU features detection
- Support C++20 standard
- algoHint parameter for some functions to allow potentially faster, but not bit-exact implementation

Core Module:
- Added int64 data type support for FileStorage
- Fixed invalid attribute value handling in FileStorage
- Extended LUT for FP16 support
- Fixed stdDev tail filling with zeros with HAL in meanStdDev
- Set and check allocator pointer for all cv::Mat instances
- Improved accuracy of Rect::contains
- Fixed result offset in minMaxIdx with HAL in some cases
- Replaced C++ operators with wrapper functions on universal intrinsics backends
- Extended cv::TickMeter
- Rewrote OpenCL-OpenGL-interop device discovery routine without extensions and with Apple support
- Marked cv::Mat(Mat&&) as noexcept
- Multiple Eigen library interop improvements

Calib3d module:
- Multiple chessboard detector improvements
- Enabled checkerboard detection with a central / corner marker on a black tile
- Fixed Rodrigues CV_32F and CV_64F type mismatch in projectPoints
- Added fisheye::distort with non-identity projection matrix
- SQPnP solver updates
- Fixed vector access in USAC

Imgproc Module:
- Added a new function that approximates the polygon bounding a convex hull with a certain number of sides
- Added Weighted Hough Transform
- Fixed bug in contours approximation
- Fixed bug in divSpectrums
- Fixed result buffer overflow in intersectConvexConvex_ for non-convex input
- Added flag to GaussianBlur for faster but not bit-exact implementation
- Added flag to cvtColor for faster but not bit-exact implementation
- Fixed fillPoly drawing over boundaries

DNN Module:
- [GSoC] Blockwise quantization support
- Faster implementation of blobFromImages for cpu nchw output
- DNN optimization with RISC-V RVV
- Added DepthToSpace and SpaceToDepth
- Yolo v10 support and related samples
- Parallel implementation of nary elementwise operations
- Support for Unflatten operation required by Attention layer
- Erf and GELU layers optimization
- Activations optimization with v_exp
- Fixed compilation errors with different OpenVINO versions
- Fixed matmul crash with CuDNN
- Fixed CuDNN runtime version check for CuDNN 9+
- Added ONNX TopK
- Fixed buffer allocation in einsum (fixed random crash on 32-bit platforms)
- Added Leaky RELU support for TFLite
- Switched to run-time dispatcher for Winograd

Objdetect module:
- Properly check markers when none are provided
- Fixed invalid vector access in QR decoder and encoder

Highgui module:
- Added new Highgui backend on top of Framebuffer
- Fixed HWND_TOP window handling on Windows
- [GSoC] Added OpenGL support with GTK3
- Several OpenGL related fixed on Linux
- Fixed leak in cvGetWindowRect_COCOA

Imgcodecs module:
- [GSoC] New API for Animations with WEBP, AVIF and Animated PNG support
- [GSoC] Add GIF decode and encode for imgcodecs
- Added experimental JPEG XL (jxl) codec support
- Initial RGB layout support in imread and imdecode
- Fixed imread output type for some cases
- Fixed file descriptor leak in HDR decoder
- Fixed corrupted JPEG decoding
- Improved error handling in image codecs
- Fixed Bayer2Gray SIMD for U8
- Avoid uninitialized value read in AVIF
- Implemented imencodemulti()
- Fixed compatibility with different OpenEXR versions

VideoIO module:
- Added VideoCapture constructor for in-memory data stream
- Fixed bugs in native video encoding on Android
- Added BGRA streams support in GStreamer back-end
- Updated materials for Orbbec cameras support
- Fixed cv::VideoWriter with FFmpeg timestamps encapsulation
- Fixed memory leak in Dshow back-end
- Fixed V4L NV12 color conversion
- Android native camera feature enhancements
- AndroidMediaNdkCapture pixel format enhancement
- Fixed VideoCapture fails to read single image with digits in name
- Fixed writer setProperty with FFmpeg plugin

Video module:
- Fixed VitTrack in the case where crop size grows until out-of-memory when the input is black
- Features2d:
- Fixed out of bounds access in SIFT

G-API module:
- Fixed input buffer read overflow in vectorized G-API convertTo implementation
- Extended G-API onnx::Params to pass arbitrary session options
- Handling I32/I64 data types in G-API ONNX back-end
- G-API: Introduce level optimization flag for ONNXRT backend

Optimizations:
- New FastCV-based HAL for Qualcomm SoCs (-DWITH_FASTCV=ON CMake option)
- Added own vectorized version of v_exp
- KleidiCV HAL for ARM updated to KleidiCV 0.3
- Initial version of HAL for RISC-V RVV 1.0 and RISC-V RVV 0.7.1 extensions
- Used LMUL=2 in the RISC-V Vector (RVV) backend of Universal Intrinsic.
- More functions in NDSRVP HAL for RISC-V P extension
- Updated built-in IPP to version 2021.12. Fixed build issues with old and new Intel IPP layouts

Platforms:
- HWAsan support on Android
- Several CUDA fixes for old GPUs without FP16 support
- Added getStdAllocator() to cv::cuda::GpuMat
- Updated NPP calls to use the new NppStreamContext API if available
- More convenient GpuMatND constructor
- Added run-time GPU check to haveCUDA
- Add support for QNX

OpenCV.js:
- Extended API white-list and added more tests
- Split white-list per module. Added opportunity to cover opencv_contrib modules
- Fix incorrect string format in js build script
- Emscripten build fixes with SIMD intrinsics
- Added more public types for USAC support
- Rename Mat::clone binding because it is used in Emscripten
- Fixed C preprocessor stringification
- Fix enum generation issues
- Multiple test improvements