Zenodo (CERN European Organization for Nuclear Research) · 2016 · 29 citations · 0 references
ComputingComputer ScienceComputer GraphicsComputer VisionDigital Imaging3D ReconstructionDepth EstimationDigital Image ProcessingImage CompressionImage ProcessingStereo VisionBug LocalizationDepth Map EstimationDepth Map ReconstructionHardware Acceleration SupportImage Quality BenchmarkingRgb-d ReconstructionScene Depth ReconstructionVideo-based Depth Estimation
This release provides performance improvements in JPEG and depth decoding with various implementations which enable hardware acceleration, including CUDA, VA-API, VideoToolbox, and Tegra. See also this performance benchmark. OpenNI2 integration is included in this release. Various internal performance improvements, error handling, and bug fixes are also included. For programmers, the API is expanded to provide Freenect2Device::startStreams() (selectively start RGB or depth stream), Registration::undistortDepth() (rectification only without registration), and Registration::getPointXYZ() (3-D points without color).