NPP Filter

Authors: Holoscan Team (NVIDIA)
Supported platforms: x86_64, aarch64
Language: C++
Last modified: October 9, 2025
Latest version: 1.0.0
Minimum Holoscan SDK version: 0.6.0
Tested Holoscan SDK versions: 0.6.0
Contribution metric: Level 1 - Highly Reliable

The npp_filter operator uses NPP to apply a filters to a Tensor or VideBuffer.

holoscan::ops::NppFilter

Operator class to apply a filter of the NPP library to a Tensor or VideBuffer.

Parameters

• filter: Name of the filter to apply (supported Gauss, SobelHoriz, SobelVert)
• type: std::string
• mask_size: Filter mask size (supported values 3, 5, 7, 9, 11, 13)
• type: uint32_t
• allocator: Allocator used to allocate the output data
• type: std::shared_ptr<Allocator>

Inputs

• input: Input frame data
• type: nvidia::gxf::Tensor or nvidia::gxf::VideoBuffer

Outputs

• input: Output frame data
• type: nvidia::gxf::Tensor or nvidia::gxf::VideoBuffer