NVIDIA C++ 2D Image and Signal Processing Performance Primitives (NPP_Plus)

NPP_Plus Documentation

The NVIDIA C++ 2D Image and Signal Processing Performance Primitives (NPP_Plus) library is a standalone, GPU-accelerated library for image and signal processing.

It provides a rich collection of highly optimized operations for performing common tasks such as filtering, segmentation, distance transforms, morphological operations, and region labeling—all tailored to take full advantage of NVIDIA GPU hardware.

Modern computer vision and image analysis applications often require multi-stage processing pipelines that involve loading, transforming, analyzing, and post-processing large volumes of data. When executed solely on the CPU, these operations can become a bottleneck, limiting the scalability and responsiveness of vision pipelines—especially in real-time and large-scale workloads.

NPP_Plus addresses these challenges by offloading intensive image and signal processing tasks to the GPU.

Built on top of CUDA, NPP_Plus provides high throughput, low-latency GPU primitives with support for batch execution, stream contexts, and multi-GPU acceleration.

In contrast to traditional CPU-based solutions or fragmented GPU utilities, NPP_Plus delivers a portable, unified solution that integrates seamlessly into modern AI pipelines, classical image processing workflows, and CUDA-based applications.

With a consistent API, simplified operator set, and support for modern C++, NPP_Plus enables fast prototyping and production deployment with minimal overhead.

Explore the documentation to learn more about using NPP_Plus for your image and signal processing needs.