The NPP (short for, NVIDIA Performance Primitive) library is a default CUDA library with a set of GPU accelerated processing functions that focus on imaging and video processing. While it enables flexible development in these fields, the developers can save their application development time. 

The NPP library has two functional parts: imaging-processing APIs, and signal-processing APIs. The image-processing APIs include tools relating to image filtering, compression/decompression, color transformation, resizing, color conversion, statistical operations, and so on. The signal-processing APIs are filtering, conversion, and so on. You can visit the NPP's document (https://docs.nvidia.com/cuda/npp), and see its configurations and the full list of functionalities.

CUDA provides many NPP-based samples. In this section, we will cover the basic use of the NPP library and discuss its application.