Alternatives
🚧 Documentation under development
The PVA ISP for NVIDIA Jetson guide is currently under active development. Some sections may be incomplete or change without notice.
Questions? Contact RidgeRun or email to support@ridgerun.com.
Alternatives to RidgeRun PVA ISP
Historically, Jetson developers have relied on two primary solutions:
- The integrated NVIDIA ISP through the Argus camera stack.
- Software-defined ISP pipelines accelerated on GPU with CUDA.
Both approaches have been successfully deployed in production systems and continue to be widely used across embedded vision applications.
This page compares the most common ISP architectures available on NVIDIA Jetson platforms and discusses the trade-offs associated with each approach.
CUDA ISP
CUDA ISP is best suited for applications that require maximum performance and flexibility in image-processing implementations.
By executing image processing algorithms directly on CUDA-enabled hardware, developers can build highly customized ISP pipelines while leveraging the performance of the Jetson GPU.
CUDA ISP is RidgeRun's software-defined ISP framework accelerated by NVIDIA CUDA.
Unlike fixed-function ISP implementations, CUDA ISP executes image processing algorithms directly on the GPU, providing developers with full visibility into the image processing pipeline and the ability to customize ISP behavior.
CUDA ISP supports image processing operations such as:
- Debayering
- Auto White Balance
- Bayer image processing
- Custom image enhancement algorithms
Because image processing executes on CUDA-enabled GPU hardware, CUDA ISP can deliver very high throughput while maintaining the flexibility of a software-defined architecture.
Advantages include:
- Full ISP visibility (due to access to the source code)
- Custom algorithm development (stages can be selected during pipeline construction)
- Software-defined image processing
- Direct GPU acceleration
The primary trade-off is that image processing workloads share GPU resources with AI inference, rendering, and other CUDA workloads.
Integrated Argus ISP
The NVIDIA camera stack provides image processing functionality through the integrated ISP and the LibArgus framework.
This solution is the default image processing path for many Jetson camera deployments and provides a mature, production-ready ecosystem for camera bring-up and image quality tuning.
Advantages include:
- Production-ready camera stack
- Mature ecosystem
- Optimized image quality
- Minimal development effort
However, the integrated ISP typically provides limited visibility into internal processing stages and limited opportunities for application-specific customization. For many production camera systems, these characteristics make Argus ISP the preferred solution.
Applications that require deeper visibility into image processing behavior, custom algorithm development, or software-defined image processing may benefit from alternative approaches.
ISP Architecture Comparison
The following table compares the main characteristics of the three ISP approaches available on NVIDIA Jetson platforms and can help determine which architecture best fits a given application.
| Feature | Integrated Argus ISP | CUDA ISP | PVA ISP |
|---|---|---|---|
| Development Effort | Low | High | Medium |
| Production Readiness | High | Application dependent | Application dependent |
| ISP Visibility | No | Yes | Yes |
| Software-Defined ISP | No | Yes | Yes |
| Algorithm Customization | No | Yes | Yes |
| Custom ISP Stages | No | Yes | Yes |
| Access to Intermediate Results | No | Yes | Yes |
| Pipeline Extensibility | No | Yes | Yes |
| GPU Utilization | None | Medium | None |
| PVA Utilization | No | No | Yes |
| Dedicated Vision Hardware | Yes (ISP) | No | Yes (PVA) |
| AI Resource Contention | Low | Medium | Low |
| GPU Resources Reserved for AI | Yes | No | Yes |
| Multi-Camera Scalability | High | Application dependent | High |
| Application-Specific Tuning | Limited | High | High |
| Power Efficiency | High | Application dependent | High |
Performance depends on image resolution, enabled ISP stages, frame rate, output format, and the number of active camera streams. Measured throughput and latency data are available in the Performance Metrics section:Performance metrics
