
CUDA ISP (libcudaisp) – GPU-Accelerated Image Signal Processor for NVIDIA Jetson & x86
CUDA ISP (libcudaisp) is a GPU-accelerated Image Signal Processing (ISP) library that brings high-performance, low-latency image enhancement to NVIDIA Jetson and x86 platforms. Perfect for embedded vision, robotics, and AI systems, this solution offloads compute-heavy ISP operations—like demosaicing, white balance, and color correction—onto the GPU, freeing CPU resources for mission-critical tasks.
If your system lacks a hardware ISP, you don't want to use the built-in ISP or is CPU-bound due to real-time vision workloads, libcudaisp gives you an option fully compatible with GStreamer pipelines.
Key Features:
-
CUDA-Accelerated ISP Functions
Includes GPU-optimized demosaicing (debayering), auto white balance (AWB), and image shifting algorithms. -
Zero Dependency on Hardware ISP
Ideal for platforms without built-in ISP (e.g., x86 + discrete GPU), or as a Jetson ISP fallback when more flexibility is needed. -
Real-Time Performance
Designed for high-resolution, real-time applications with low latency requirements. -
GStreamer Integration
Easily integrates into multimedia pipelines using RidgeRun’scudadebayer
,cudaawb
, andcudashift
plugins. -
Modular and Extensible
Built as a C++ library, making it easy to extend with custom image processing modules.
Use Cases:
-
Embedded AI applications requiring clean, corrected video input
-
Robotics systems needing real-time GPU ISP on the edge
-
Streaming platforms offloading ISP from CPU to GPU
-
Vision systems running on Jetson Xavier NX, Orin, or x86 with CUDA GPUs
Supported Platforms:
-
Jetson Nano, Jetson Xavier NX, Jetson AGX Orin
-
x86-64 Linux with NVIDIA GPU (e.g., for desktop inference or cloud edge)
What You Get:
-
Full access to RidgeRun’s CUDA ISP (libcudaisp).
-
GStreamer plugin support (optional):
cudadebayer
,cudaawb
,cudashift
-
Documentation, sample pipelines, and quick-start guides
-
Optional Embedded Software Support package for integration, tuning, and extensions
Why Choose RidgeRun?
With over 19 years of experience in Embedded Software Consulting, RidgeRun delivers production-quality software for demanding real-time systems. libcudaisp is backed by our world-class engineering team, ensuring your image pipeline meets both performance and quality benchmarks.
Need a complete imaging pipeline, from camera drivers to AI inference? RidgeRun’s subscription-based development model is here to help.
See full technical documentation
Read our blog overview