CUDA MJPEG codec for NVIDIA GPUs

Fastvideo CUDA Motion JPEG Codec offers high speed compression and decompression of MJPEG streams on GPU. We got ultra fast MJPEG compression and decompression on the GPU due to full parallel implementation of encoding and decoding algorithms.

For high speed imaging applications MJPEG is the only option for realtime encoding. H.264/H.265/AV1 codecs could offer better compression ratio, but they are unable to work at frame rates more than 200 fps. With MJPEG one could achieve video encoding at 1000 fps and more.

Features of Motion JPEG Codec on CUDA

• Grayscale RGB8 MJPEG streams
• Color RGB24 MJPEG streams with 4:2:0, 4:2:2 and 4:4:4 subsampling formats
• Realtime image processing on GPU
• Compatibility with FFmpeg to read/write MJPEG streams (FFmpeg is under LGPLv2.1)
• System requirements: Windows-10, Linux Ubuntu, L4T

Options for CUDA MJPEG video compressor

We can offer custom software design for combined debayer and JPEG encoder on NVIDIA GPU and for the following solutions:

• Debayer + MJPEG encoder
• Image preprocessing + Debayer + Crop + Resize + Sharp + OpenGL output + JPEG encoding + FFmpeg

We can also extend our MJPEG codec with fast image processing pipeline for high speed and high resolution cameras: image acquisition, dark frame subtraction, flat-field correction, white balance, noise removal, demosaicing, image filtering, color correction, 1D/3D LUT, tone mapping, cropping, resizing, remapping, sharpening, mjpeg compression on GPU, etc.

Roadmap for further improvements of CUDA MJPEG Codec

• 12-bit CUDA JPEG codec
• Minimum memory usage on GPU
• More input/output formats for MJPEG codec (YCbCr, YV12, etc.)
• MJPEG streaming in realtime (mjpeg video codec on cuda)
• MJPEG codec on GPU for FFmpeg
• Integration into GPU RAW Processor

Support

• Full technical support up to successfull client integration
• SDK, documentation
• Fastvideo SDK benchmarks

For any further information concerning CUDA MJPEG codec please contact us via email or contact form.