FastVCR Low-Latency Software on GPU for Hikrobot Cameras

Updated: August 27, 2026

FastVCR brings a complete GPU-based ISP to Hikrobot machine-vision cameras: raw frames arrive over USB3 or GigE, and the entire processing chain — from unpacking and linearization through demosaicing, denoising, and tone mapping to encoding — runs on the NVIDIA GPU, with output to the display, to JPEG or MP4 recordings on SSD, or to a low-latency RTSP stream. Hikrobot is a global manufacturer of machine-vision products for industrial automation (positioning, measurement, and quality inspection). A demo version for your Hikrobot camera is available upon request — use the contact form below.

How these numbers were measured. Fastvideo benchmarks are measured on NVIDIA GPUs, timing pure GPU processing at the resolution, bit depth, format and quality stated on this page. See the full benchmark methodology and test environment and the Fastvideo SDK benchmark report (PDF). Our tools are open source on GitHub and built on NVIDIA CUDA.

Check a single module. Every module in the Fastvideo SDK comes with a command-line sample. Run it on your own images on your own GPU and compare the result with the figures published here.

Check the whole pipeline. A module in the middle of a pipeline cannot be judged on its own: its input is hard to prepare, and its output is an intermediate product you cannot look at. What matters is how it changes the final image and the total time of the schema. Two of our applications run the complete pipeline and report the time of every stage — FastVCR (live cameras or RAW files from disk, so no camera is required) and Fast CinemaDNG (DNG, offline). The benchmark module is included in the free builds of both.

Detailed benchmark data. Measured results are published on our benchmark pages: all benchmarks.

We wrote these algorithms ourselves

Image processing has always been a trade-off between quality and speed. The methods that give a good result take minutes per frame; the fast ones leave halos and artifacts. The reason is the compute budget: on a CPU there was never enough of it, so algorithms were simplified — and much of the industry still runs on simplifications accepted decades ago.

Fastvideo’s codecs and ISP modules are neither ports of that legacy nor wrappers around third-party code. We designed — and in many cases invented — the algorithms behind our demosaicing, denoising, tone mapping and JPEG/JPEG2000 codecs, and wrote them for CUDA from the start, on top of image-processing expertise built since 2009. An algorithm moved to the GPU still carries the compromises it was simplified for; ours were never made under that constraint. On a modern GPU the budget is large enough that, for practical workloads, the choice between quality and speed no longer has to be made.

That is why you get a complete, tuned imaging solution rather than a bag of GPU primitives — and why, when your pipeline needs something specific, we can change the algorithm itself, not just call it.

The Gap Between Camera Bandwidth and CPU Processing

Sensor progress has outpaced host-side processing. Current Sony Pregius, Gpixel, and AMS (CMOSIS) sensors ship in high-speed variants where resolution and frame rate grow together, and Hikrobot pairs them with USB3 and GigE interfaces capable of delivering every captured frame to the host. Decoding, demosaicing, and enhancing that stream in software is where CPUs fall behind: frames get dropped or quality gets sacrificed. The vendor's MVS application is a capable configuration and testing tool, but it was never meant for sustained high-fps processing. FastVCR moves the whole workload to the NVIDIA GPU — from mobile to server class — so the full-quality pipeline, including proper color science, keeps pace with the camera at its native bandwidth.

Camera Setup and 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, etc.

GPU Image Processing Pipeline

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: wavelet-based, bilateral, NLM
Curves and levels
Local tone mapping, including shadow recovery
Sharpening (local contrast)
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

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

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

Real-time Monitoring Tools

Histograms for RAW and processed data, RGB parade, vectorscope
Camera statistics
Time measurements for every image processing module on GPU
Glass-to-glass (G2G) module for system latency estimation

Performance

How does FastVCR keep up with high-bandwidth Hikrobot USB3 and GigE cameras?

FastVCR is built for cameras that saturate their interfaces. With USB3 and GigE models it handles multi-camera setups, processing raw streams from all connected cameras simultaneously without frame drops; the highest throughput is reached with 10-GigE cameras, where the interface bandwidth allows the largest data rates. As with any ISP, the achievable numbers depend on the pipeline configuration and the hardware (CPU/GPU/SSD), so the most reliable benchmark is your own: the demo runs the same pipeline as the full version.

A camera emulator mode is included for evaluation without hardware: point the software at a folder of RAW/PGM/TIFF frames on SSD, and it will process them sequentially as a virtual camera, with per-module GPU timing available for profiling.

Latency

FastVCR measures end-to-end delay with a built-in glass-to-glass (G2G) tool. A live timer on screen is captured by the Hikrobot camera, run through the full GPU pipeline, and drawn next to the original — the gap between the two readings is the system latency; a second mode repeats the test through MJPEG or H.264/H.265 encode and decode, locally or over the network. Because the figure moves with frame rate, monitor refresh, GPU, link speed, and pipeline depth, treat any published number as guidance and measure your own automation cell with the demo.

FastVCR vs MVS: What the GPU Adds

Hikrobot’s MVS covers configuration and testing; FastVCR is the processing layer for automation and inspection lines that have to run at full frame rate over USB3, GigE, and 10-GigE.

Throughput limited by the GPU, not the host. The full 16/32-bit ISP with MG demosaicing executes on the GPU, so an inspection station keeps pace at the camera’s native bandwidth instead of stalling a CPU thread.

Encoding that matches line speed. GPU MJPEG records whatever resolution and fps the sensor emits, past the point where H.264/H.265 encoders give up, so capture never gates the line.

Deterministic, measurable latency. A built-in glass-to-glass test reports real numbers for your automation cell rather than a datasheet figure.

Denoising for demanding inspection. Wavelet, bilateral, and NLM denoisers for monochrome and color imaging.

Validate before deployment. Emulator mode replays RAW/PGM/TIFF sequences from SSD, so a pipeline can be proven out without stopping the line.

Built for unattended runs. A remotely controllable CLI sits beside the GUI, both with an RTSP server and player — suited to aerial imaging, street-view capture, and long recordings.

Shop-floor monitoring. Real-time histogram, parade, vectorscope, and detailed camera statistics.

Optional Modules on GPU

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, XIMEA, Emergent Vision, IO Industries, Daheng Imaging, MindVision and other cameras

FastVCR Command-Line Application

Alongside the GUI, FastVCR ships as a CLI application with the identical feature set — for headless hosts, UAV payloads, remote camera control, and unattended long-term recording or streaming. Sensor and processing parameters remain fully adjustable at runtime; for preview, use the bundled RTSP player or VLC. The CLI runs on Windows, Linux, and L4T with all processing on the GPU.

Under the hood, FastVCR is an application built on the Fastvideo CUDA image processing SDK — the same GPU modules can be licensed separately and embedded into your own software. We also take on custom development: application-specific processing modes and complete solutions to your specification. 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
Hikrobot USB3 and GigE cameras
Processing of 8–16-bit RAW/PGM/TIFF images from SSD (camera emulator mode)

Software Downloads

FastVCR demo for Hikrobot cameras — available upon request via the contact form below
PDF manual for FastVCR software

Applications

Machine vision and industrial automation
Online visual inspection and quality control
Aerial imaging and drone photography
Street-view capture and mobile mapping
Ophthalmology, retinal surgery, and medical imaging
Broadcasting and streaming
3D scanning and robotic arms
Data collection for AI applications with machine vision cameras

Frequently asked questions

What is FastVCR for Hikrobot cameras?

FastVCR turns a Hikrobot USB3 or GigE camera into a fully GPU-processed imaging system: raw frames are decoded and run through the complete ISP — demosaicing, denoising, color and tone — on an NVIDIA GPU, then shown on screen, written to JPEG/MP4 on SSD, or streamed over RTSP. It packages the Fastvideo SDK as a ready-to-run camera application.

Which Hikrobot cameras work with FastVCR?

Hikrobot machine-vision models on USB3 and GigE, up to 10-GigE, in monochrome or Bayer at 8–16-bit. Multiple cameras can be captured at once; 10-GigE links carry the highest data rates and give the most processing headroom.

How is FastVCR different from Hikrobot MVS?

Hikrobot’s MVS is a solid configuration and testing tool but is not designed for sustained high-fps processing. FastVCR performs full-quality 16/32-bit ISP on the GPU, so the limit is the GPU and the pipeline you choose rather than the CPU, and GPU MJPEG recording keeps pace with the sensor.

How do I check performance and latency on my setup?

FastVCR includes a built-in glass-to-glass module and per-module GPU timers. Because the free demo runs the exact same pipeline as the licensed build, the most reliable figures are the ones you measure on your own Hikrobot hardware — see the benchmark methodology.

Can I evaluate FastVCR without a Hikrobot camera?

Yes. Emulator mode reads RAW/PGM/TIFF sequences from SSD and feeds them through the pipeline as a virtual camera — handy for profiling and regression testing before the hardware arrives.

FastVCR Roadmap

FastVCR evolves alongside the Fastvideo SDK with regular updates; the list below shows the latest additions relevant to Hikrobot cameras (through 2026) and what is currently in development.

Released:

Adaptive local tone mapping (ALTMapper) engine for single shot HDR
NLM and bilateral denoisers on the GPU
High quality color calibration for camera applications
CUDA-12.6 support for Windows/Linux/L4T

In progress:

Chromatic aberration removal in the RAW domain with self-calibration
Compatibility with YOLO
RAW denoising

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.