Xilinx is now part ofAMDUpdated Privacy Policy

Computer Vision and Image Processing are ubiquitous today in a wide range of applications like Medical Imaging, ADAS, Robotics, IIoT, Surveillance, and Video Streaming services and are also a critical part of the end-to-end processing pipeline of AI-powered vision solutions. These applications demand a solution that can meet real-time performance and flexibility to manage a range of frame resolutions and adaptable throughput requirements (1080p60 to 8K60), while being power-efficient. The architecture of Xilinx platforms provides the ideal solution to meet your vision system requirements, both at the edge and in the data center.

Vitis™ Vision library enables you to develop and deploy accelerated computer vision and image processing applications on Xilinx platforms, while continuing to work at an application level.  These library functions offer a familiar interface like the OpenCV libraries and have been optimized for performance and resource utilization on Xilinx platforms. They also offer the flexibility to meet the adaptable throughput requirements of your vision systems.

Key Features:

  • Performance-optimized kernel and primitive functions for:


    • Color and bit-depth conversion, Channel Extractions, Pixel-wise Arithmetic Operations
    • Geometric Transforms, Image Statistics, Filters
    • Feature Detection and Classifiers
    • 3D Reconstruction
    • Motion Analysis and Tracking


  • Support for color image processing and multi-channel support
  • Efficient management of data movement between on-chip or external memory for best performance
  • Function parameters enable processing multiple pixels/clock to meet throughput requirements
  • Several design examples demonstrate how to accelerate your vision and imaging algorithms step-by-step

Performance Benchmark

Stereo Local Block Matching

Vitis Vision Library Vs. OpenCV

132x Speed Up

Frame Size
1242 (W) x 375(H)
CPU (1) 10 Frames/Sec
Xilinx Alveo U200 (2) 1320 Frames/Sec


  • 12 Parallel Streams on Xilinx Alveo U200 @300MHz
  • Only 37.35% Total Device Utilization 
  • Latency: 9.7 ms

Getting Started