Fast VCR Low-Latency Software on GPU for XIMEA Cameras

Updated: August 4, 2026

FastVCR (Fast VCR) is a camera application with a complete image processing pipeline on NVIDIA GPU, designed for XIMEA USB3 and PCIe cameras. It captures raw frames from the camera, performs the full ISP on the GPU — from linearization and demosaicing to denoising, tone mapping, and encoding — and outputs the result to the monitor, to JPEG or video files on SSD, or to a low-latency RTSP stream. On the NVIDIA GeForce RTX 4090 it processes a 65 MPix camera stream in real time, and the pipeline is designed for low glass-to-glass latency. You can download the demo and test it with your XIMEA camera (8/10/12-bit raw, monochrome or Bayer).

Working with a specific XIMEA line or CamTool? See the xiB PCIe cameras, the xiQ USB3 cameras, or the XIMEA CamTool GPU plugin. This page is the FastVCR application for XIMEA cameras across USB3 and PCIe.

Why GPU Software for Modern Cameras

Modern machine vision cameras combine high resolution with high frame rates: the latest generations of Sony Pregius, Gpixel, and AMS (CMOSIS) image sensors, paired with fast USB3 and PCIe interfaces, deliver raw data at rates that CPU-based processing cannot handle in real time without dropping frames. The vendor's CamTool software is fine for camera testing, but not for real-time processing at full bandwidth. FastVCR closes this gap: all image processing runs on the NVIDIA GPU (mobile, laptop, desktop, or server), so the pipeline keeps up with the camera at maximum resolution and frame rate — with high image quality and proper color science, not just raw speed.

Supported XIMEA Cameras

xiB-64 series — high performance PCIe cameras with up to 64 Gbps bandwidth
xiC series — compact USB3 cameras with Sony Pregius sensors
xiMU series — miniature USB3 cameras, the smallest industrial grade USB cameras
xiQ series — USB3 Vision standard cameras with single board design
xiX series — board-level PCIe cameras for embedded vision and multi-camera systems

Camera Control Software

Camera initialization: camera model, default values for frame resolution, fps, etc.
Input bit depth and pixel format
Image resolution and region of interest (ROI)
Frame rate (fps), exposure time, gain, camera bandwidth
Automatic exposure/gain control with overexposure suppression
Software presets for fast change of image sensor parameters
Optional sensor modes: binning, decimation, dual gain, dual ADC, etc.
XIMEA camera lens control

Image Processing Pipeline on GPU

Image acquisition and frame unpacking for 10-bit and 12-bit modes
Image linearization
Dark frame subtraction (FPN)
Flat-field correction (shading correction)
Bad pixel removal
White balance / AWB
Adaptive exposure and gain control
High quality demosaicing with the MG algorithm
Color correction with matrix profile or DCP profile
Highlight recovery
Exposure correction (brightness control)
Denoising: bilateral, NLM
Local tone mapping, including shadow recovery
Sharpening
Gamma transform
Crop, resize (downscale and upscale)
Rotation to 90/180/270 degrees and flip/flop; rotation to an arbitrary angle
Undistortion via LCP or via calibrated maps

Recording, Direct-to-Disk Capture, Streaming and Output

Real-time video output to monitor via OpenGL
JPEG compression with image storage on SSD
Video encoding to MJPEG (AVI) and H.264/H.265/AV1 (MP4) with storage to SSD
Automatic real-time partitioning of AVI/MP4 files to a specified file size
Built-in RTSP server for low-latency streaming; RTSP player included, and the stream is compatible with VLC
Real-time RAW/PGM data storage on SSD
Interoperability with third-party GPU-based SDKs and AI libraries at the GPU level

Monitoring and Analysis

Histograms for RAW and processed data, RGB parade, vectorscope
Camera statistics
Time measurements for every image processing module on GPU
Glass-to-glass module for system latency estimation

Performance Benchmarks

The reference configuration: FastVCR processes raw frames from a 65 MPix camera (9344×7000) at 70 fps in 8-bit mode or 60 fps in 10-bit mode on the NVIDIA GeForce RTX 4090. The measured pipeline includes RAW import, RAW-to-RGB transform, color correction, JPEG encoding, and storage to SSD. With USB3 cameras the software also runs multi-camera systems, processing raw frames from all cameras in real time without frame drops; the maximum performance is reached with PCIe cameras due to their bandwidth. Actual performance depends on the pipeline complexity and the hardware (CPU/GPU/SSD).

FastVCR real-time performance with XIMEA cameras on the NVIDIA GeForce RTX 4090: resolution, bit depth and frame rate.
FastVCR on NVIDIA GeForce RTX 4090	Resolution	Bit depth	Frame rate
XIMEA 65 MPix camera	9344×7000	8-bit	70 fps
XIMEA 65 MPix camera	9344×7000	10-bit	60 fps

The software also offers a camera emulator mode: you can load an arbitrary RAW image in RAW/PGM/TIFF format and measure the GPU processing performance without a camera connected. Per-module time measurements make it a convenient tool for pipeline profiling.

Latency

FastVCR is built for low glass-to-glass (G2G) latency and includes a built-in G2G module so you can measure it on your own system. For the measured figures, the G2G test methodology and the factors that affect latency, see our low-latency camera software page.

FastVCR vs CamTool: What the GPU Adds

XIMEA’s bundled CamTool is made for setup and testing; FastVCR is made to process every frame at full quality and full speed on high-resolution USB3 and PCIe sensors.

GPU-bound, not CPU-bound. A 16/32-bit pipeline with high quality MG demosaicing runs entirely on the GPU, so the ceiling is your card and the ISP you configure — not a saturated CPU core — which is what lets a 65 MPix stream stay real-time.

Recording that never throttles the sensor. GPU MJPEG absorbs resolution and frame-rate combinations that H.264/H.265 encoders refuse, so writing to SSD is never the stage that drops frames.

Latency you can verify. RTSP transport and fast RAW processing keep glass-to-glass delay low, and the bundled G2G tool lets you confirm the latency on your own workstation.

Scientific-grade denoising. Bilateral and NLM denoisers for monochrome and color, tuned for image quality rather than speed alone.

A camera-free workflow. Point the software at RAW/PGM/TIFF folders on SSD to replay them as a virtual camera for evaluation, regression tests, and pipeline tuning.

Interactive or headless. A GUI and a remotely controllable CLI ship together, each with an RTSP server and player.

Instrument-style monitoring. Live histogram, RGB parade, vectorscope, and per-module GPU timing with full camera statistics.

Optional Modules on GPU

Curves and levels for RAW, RGB, HSV
DCP support and calibration: CCM, HSV LUT, tone curve
3D LUT support
Global tone mapping
HDR support
Geometric and affine transforms
Undistortion via accelerated remap — in progress
Chromatic aberration removal in the RAW domain — in progress
SDI output support to work with Blackmagic grabbers
Multiple camera support; fast replay (synchronized viewer) for multi-camera systems
High performance JPEG2000 and Raw Bayer codecs
Interoperability with FFmpeg
Camera and lens calibration: FPN, FFC, DCP, undistortion
GenICam (GenTL) support to work with Basler, Imperx, FLIR, Lucid, Emergent Vision, IO Industries, Daheng Imaging, MindVision, Hikrobot and other cameras

Fast VCR Command-Line Application

Many tasks require running the software without a GUI: UAV applications, remote camera control, long-term unsupervised recording and streaming. For these cases we've developed a CLI application with the same feature set as the GUI version. It provides full real-time control over the image sensor and processing parameters; for video preview you can use our RTSP player or VLC. The CLI application runs on Windows, Linux, and L4T, with all image processing on the GPU.

All the above features are built on our CUDA image processing SDK, so the same GPU-accelerated components can be licensed and integrated into your own applications. Apart from FastVCR, we offer custom software development according to your requirements, including application-specific modes of operation — please describe your project in the contact form below.

Compatibility

OS: Windows 10/11, Linux Ubuntu, Linux4Tegra (L4T)
NVIDIA GPUs (GeForce RTX, NVIDIA RTX) with CUDA 12.6
NVIDIA Jetson: Nano, Xavier NX/AGX, Orin NX/AGX
XIMEA USB3 and PCIe cameras
Processing of 8/12/16-bit RAW/PGM/TIFF images from SSD (camera emulator mode)

Software Downloads

FastVCR demo for XIMEA cameras — Windows 10/11, CUDA 12.6; works with a connected XIMEA camera and with PGM/RAW/TIFF frames from SSD
FastVCR portable demo — Windows 10/11, CUDA 12.6; processing of PGM/RAW/TIFF frames from SSD only, no camera required
FastVCR for Linux Ubuntu 22.04 (CUDA 12.6) — available upon request
FastVCR for NVIDIA Jetson — available upon request
PDF manual for FastVCR software

Applications

Ophthalmology, retinal surgery, and medical imaging
Machine vision and industrial applications
Media and entertainment
Broadcasting and streaming
Drones and UAVs
3D scanning and robotic arms
Aerial imaging
Data collection for AI applications with machine vision cameras

Frequently asked questions

What is FastVCR?

FastVCR (Fast VCR) is a camera application that runs the complete image processing pipeline on an NVIDIA GPU for XIMEA USB3 and PCIe cameras — raw capture, full ISP (demosaicing, denoising, color, tone mapping), and output to monitor, to JPEG or video files on SSD, or to a low-latency RTSP stream. It is built on the Fastvideo SDK.

Which XIMEA cameras does FastVCR support?

XIMEA USB3 and PCIe models, including the xiB-64, xiC, xiMU, xiQ and xiX series, with 8/10/12-bit raw input, monochrome or Bayer. Multi-camera systems are supported, and the highest throughput is reached with PCIe cameras.

How fast is FastVCR?

On an NVIDIA GeForce RTX 4090 it processes a 65 MPix camera (9344×7000) in real time at 70 fps in 8-bit mode or 60 fps in 10-bit mode, including RAW import, RAW-to-RGB, color correction, JPEG encoding and storage to SSD. Exact numbers depend on the pipeline and your CPU/GPU/SSD — see the benchmark methodology.

What is the glass-to-glass latency?

Glass-to-glass (G2G) latency is the delay from the sensor capturing a frame to that frame appearing on the display. FastVCR is built to keep it low and includes a built-in G2G module to measure it on your own system. For measured figures and the factors that affect latency, see our low-latency camera software page.

Can I try FastVCR without a XIMEA camera?

Yes. In camera-emulator mode it processes folders of RAW/PGM/TIFF frames (8/12/16-bit) from SSD, so you can evaluate the pipeline and measure GPU performance without a camera — download the FastVCR demo (7z).

What image and video formats can FastVCR record?

JPEG for still images, MJPEG in AVI, and H.264, H.265 or AV1 in MP4 for video, plus raw RAW/PGM storage — all written to SSD in real time, with automatic partitioning of AVI/MP4 files by size. The processed stream can also be sent out over RTSP.

Can FastVCR stream video over RTSP?

Yes. FastVCR includes a built-in RTSP server for low-latency streaming, with an RTSP player included; the stream is also compatible with VLC.

Does FastVCR run on NVIDIA Jetson?

Yes. Besides desktop and server NVIDIA GPUs, FastVCR runs on Jetson (Nano, Xavier NX/AGX, Orin NX/AGX) under Linux4Tegra; the Jetson build is available on request.

Fast VCR Roadmap

FastVCR is developed on top of the Fastvideo SDK and updated regularly; the items below are its most recent releases (through 2026) and the work currently in progress.

Released:

Adaptive local tone mapping (ALTMapper) engine for single shot HDR
Low-latency software for XIMEA cameras
CUDA 12.6 support for Windows/Linux/L4T
XIMEA camera lens control
High quality color calibration for camera applications
Lossless RAW Bayer compression on the GPU
Support for Emergent Vision cameras: 10/25/50/100-GigE
Software for MIPI CSI-2 cameras with CUDA-based ISP
CLI application with remote control and long-term operation

In progress:

Chromatic aberration removal in the RAW domain with self-calibration
RAW Bayer codec on the GPU
Support for XIMEA xiB, xiX-XL, and xiX-Xtreme camera series

Blog Posts on the Subject

About the author

Fyodor Serzhenko, PhD, is the founder and CEO of Fastvideo. He earned his PhD at the Moscow Institute of Physics and Technology (MIPT) in 1993. Since 2009 he has led the development of Fastvideo’s GPU-accelerated image codecs and ISP modules, including the Fastvideo SDK. Connect on LinkedIn.

Why you can trust these results

Fastvideo has built GPU-accelerated image processing software since 2009. Our codecs comply with the standards they implement — JPEG (ITU-T T.81 / ISO IEC 10918) and JPEG2000 (ITU-T T.800 / ISO IEC 15444) — and our tools are open source on GitHub. Every performance figure we publish is reproducible: see our benchmark methodology, download the Fastvideo SDK benchmark report (PDF), and measure it on your own GPU.