What is PVA ISP?

🚧 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.

PVA ISP for
NVIDIA Jetson

Table of Contents [Sticky]

What is RidgeRun PVA ISP?

PVA ISP is RidgeRun's hardware-accelerated Image Signal Processor (ISP) accelerated by NVIDIA's Programmable Vision Accelerator (PVA), available in NVIDIA Orin and Thor, providing a configurable PVA-accelerated image-processing pipeline that transforms raw sensor data into formats suitable for display, video encoding, streaming, and computer vision applications. All this without loading the GPU with ISP tasks.

PVA ISP is an alternative to NVIDIA's Argus hardware-accelerated ISP and is adequate for other capture systems, such as the Holoscan Sensor Bridge, where Argus is no longer used. Moreover, PVA ISP enables full pipeline visibility, configurable processing stages, access to intermediate results, and the flexibility to integrate custom algorithms and application-specific workflows.

Unlike traditional ISP implementations that operate as fixed-function pipelines, PVA ISP exposes image processing stages as configurable accelerated components, enabling customization, tuning, and future extensibility while leveraging dedicated vision acceleration hardware.

Figure 1. High-level overview of the PVA ISP functionality.

As computer vision systems become more specialized, developers increasingly require:

Custom image processing algorithms
Application-specific image tuning
Access to intermediate processing stages
Easier debugging and validation
Extensible processing pipelines

PVA ISP was created to provide a flexible image processing framework capable of adapting to application-specific requirements while leveraging NVIDIA's dedicated vision processing hardware.

The current PVA ISP implementation supports the image processing stages shown below.

Stage	Purpose
Black Level Correction	Removes sensor offsets and establishes a proper black reference level.
Digital Gain	Applies configurable gain adjustments to image data before color reconstruction.
Auto White Balance (AWB)	Corrects color casts caused by varying illumination conditions.
Demosaicing	Reconstructs full RGB color information from Bayer sensor measurements.
Histogram Generation	Produces image statistics used for image analysis and tuning.
Tone Mapping	Compresses dynamic range while preserving visual detail.
Gamma Correction	Adjusts image brightness for display and human perception.
RGB to NV12 Conversion	Converts RGB images into a format commonly used by video encoders and multimedia pipelines.

For additional information about each stage you can consult at Supported PVA ISP algorithms

Common Use Cases

PVA ISP is particularly well suited for applications where image processing behavior must be understood, validated, customized, or optimized for specific operating conditions.

These applications often require greater control than what is typically available through black-box ISP implementations and may involve challenging imaging environments, application-specific algorithms, or strict validation requirements.