GPU JPEG Compression

We have created Fast GPU JPEG codec for NVIDIA GPU. We got ultra fast GPUJPEG compression and decompression due to parallel implementation of Baseline JPEG algorithm. Our GPU JPEG codec is much faster in comparison with the best commercial multithreaded JPEG codecs for multicore CPUs.

Fast JPEG codec features

  • Baseline JPEG for grayscale (8-bit) and color (24-bit) images
  • 12-bit JPEG compression on GPU
  • Very fast lossy image encoding and decoding with variable compression ratio
  • Data input: 8/24-bit images in CPU/GPU memory or HDD/SSD
  • Data output: final compressed/uncompressed JPEG image in CPU/GPU memory or HDD/SSD
  • Continuous data mode (input one image after another)
  • Standard set of computations for parallel implementation of Baseline JPEG: Color Transform, 2D DCT, Quantization, Zig-zag, AC/DC, DPCM, RLE, Huffman, Byte stuffing, JFIF formatting
  • Maximum input image size is 16,000×16,000 or even more (optional)
  • Compatibility with FFmpeg to read/write MJPEG streams
  • Optimized for the latest NVIDIA GPUs (Kepler, Maxwell, Pascal, Volta, Turing)
  • Compatible with Windows-10, Linux Ubuntu/CentOS, L4T

Benchmarks for JPEG encoding on GeForce RTX 2080TI

Now we need just 3.3 ms for Baseline JPEG encoding of 24-bit color image with 9344×7000 resolution, JPEG quality 90%, subsampling 4:2:0. That encoding time corresponds to JPEG compression performance around 60 GB/s. This is the result for single image mode (without batch and without streaming) if input/output data reside in GPU memory.

More info about GPU JPEG Codec


For any further information concerning GPU JPEG codec or free trial please contact us via email or contact form.

Contact Form

This form collects your name and email. Check out our Privacy Policy on how we protect and manage your personal data.