CUDA GPU Solution drastically increases performance of cameras

Software collaboration

XIMEA always strives to offer to customers something extra on top of everything else.

We call it - To go the extra Vision.

This is why cooperation with company called Fastvideo was so important for us.

You can explore the new possibilities below.

Latest audio/video sessions from GTC 2014 Conference on the topic:
 

Fastvideo Image Processing SDK

Fastvideo company has designed high performance SDK for image processing on GPU with peak performance able to reach 1.2 GPix/s.

For 12-bit image with resolution 5120x3840 it takes under 16 ms on NVIDIA GeForce GTX 980 GPU.

This SDK corresponds to standard image processing pipeline for camera applications.

It covers all image processing stages starting from RAW image acquisition from a camera to JPEG compression with storage to RAM or SSD/HDD.

All image processing can be done on single GPU and this leads to exceptionally high performance.

Image & Video Processing SDK Features

• Image acquisition module from a camera
• Unpacking module for various RAW formats
• Dark frame subtraction (FPN)
• Flat-Field Correction (Shading Correction)
• White Balance
• Exposure correction
• RAW denoiser
• High quality debayer
• Color noise removal
• Color correction with matrix
• RGB and HSL tone mapping with LUTs (curves and levels)
• Rotation, Crop, Resize, Sharp
• LUT for 16-to-8-bit transform
• RGB24 Histogram and Parade (waveform monitor)
• Scaling and output to monitor via OpenGL
• JPEG compression and decompression
• Writing of compressed stream to AVI

 Image & Video Processing SDK Options

• Auto White Balance
• Discrete Wavelet transforms 5/3 and 9/7
• Medial filter
• Time measurements for all SDK modules
• Trace module
• Multiplexor module
• Support for input data from HD-SDI grabbers: Bluefish and Deltacast
• Multiple camera solutions for real-time processing
• NVIDIA Tegra K1, X1 and X2 support
• Stream-per-thread option for better performance
• Interoperability with third-party SDKs on CUDA
• Interoperability with FFmpeg
• RAW and DNG (CinemaDNG) image and video processing

Fast compression and realtime image and video processing on GPU is popular and growing in demand in modern industries.

Indeed, high framerate cameras, real-time video surveillance systems, heavy-loaded web services and medical equipment often require zero tolerance to delays due to image processing. And this is where the image compression or image processing speed can become a serious bottleneck.

Comparing with classic CPU solutions, Fastvideo's GPU algorithms demonstrate 10-30 times performance speed up retaining the same or even better quality of the output and minimizing the host processor workload.
 

Compatibility

• OS Windows-7/8/10, Linux Ubuntu and CentOS, L4T
• NVIDIA GPUs (Kepler, Maxwell, Pascal, Volta) for mobile, laptop, desktop, server
• NVIDIA Tegra K1, X1 and X2 mobile GPUs
• NVIDIA CUDA 9.2
• Sample projects for MSVC 2017 and gcc

Custom Image Processing SDK for NVIDIA GPUs

Wide range offering of integration services to connect your software with Fastvideo's image and video processing SDK.

To achieve high performance on GPU for specific imaging applications Fastvideo is also developing custom SDKs according to agreed specification.

If you can't find in the details above your image processing pipeline, get in touch to discuss overall possibility.

You can also have a look at our projects in development.

Inspired? What’s Next?