JPEG2000 lossless 12-bit RAW encoding on Jetson Orin AGX

Updated: July 23, 2026

Fastvideo's JPEG2000 codec performs lossless 12-bit raw encoding on NVIDIA Jetson AGX Orin at up to 650 MPix/s, fast enough for on-camera compression in real-time embedded systems. This makes lossless RAW archiving practical directly on the Orin module.

Applications for real-time lossless J2K raw encoding on Jetson Orin

Street-view
UAV
Medical imaging
Microscopy
Digital pathology
Geospatial imaging
Digital cinema
Satellite imagery

It's worth mentioning that a lossless encoding algorithm is often essential for these applications. We usually need to process 12-bit raw data from monochrome, infrared or Bayer cameras. The final RAW-to-RGB or RAW-to-Grayscale conversion can be done later; for now, we just need to capture and to compress the input raw frames using a lossless algorithm, and that should be done at real time.

You can find detailed information about our testing approach and the latest benchmarks for JPEG2000 encoding and decoding on desktop/server NVIDIA GPUs you can find at the corresponding links. The procedure is exactly the same, but it is applied to the NVIDIA Jetson Orin hardware.

lossless JPEG2000 raw encoder Jetson Orin

JPEG2000 lossless raw encoding parameters

Input file format – PGM
Output file format – JP2
Bit depth: 12 bits
Test image resolution: 5312 × 4608
Lossless JPEG2000 compression with CDF 5/3 wavelet
Number of DWT resolutions – 6
Codeblock size – 32×32
MCT – on
PCRD – off
Tiling – off
Window – off
Quality layers – one
Progression order – LRCP (L = layer, R = resolution, C = component, P = position)
Modes of operation – single, loop or multithreaded batch
24 MPix 12-bit raw bayer RGGB image – RGGB_5312x4608_12bit.pgm
Compression ratio for JPEG2000 lossless encoding - 2.8

Our lossless encoding benchmarks for that particular image are intended to demonstrate the level of performance that can be achieved with Jetson Orin AGX hardware and our GPU implementation of JPEG2000 lossless algorithm. However, as image compression ratio is content-dependent, you may get different results for your images. Nevertheless, the encoding performance level should remain consistent.

Clearly, the compression ratio for lossy (visually lossless) encoding could be much higher for the JPEG2000 algorithm than for exact lossless compression. Therefore, we recommend testing different parameters to achieve the best compression ratio with acceptable image quality. Decreasing the quality coefficient not only improves compression, but also increases the performance both for encoding and decoding. Our benchmarks show the performance results for the above images and parameters.

Hardware and software

NVIDIA Jetson Orin AGX, Jetpack 6.2
CUDA Toolkit 12.6

JPEG 2000 encoding benchmarks on NVIDIA Jetson Orin AGX for 12-bit 24-MPix RAW image

Hardware	1 image	100 times average	2 threads, 100 times
JPEG2000 Encoder on Jetson Orin AGX	6.7 fps	13.2 fps	27 fps

As we can see, we've achieved JPEG2000 lossless encoding performance of around 650 MPix/s for that particular 12-bit raw image at a 2-thread application. This is much higher than the performance requirements for various applications in a project.

Please note that we can also offer lossy JPEG2000 raw encoding on Jetson Orin. This solution provides superior performance and much better compression ratio, albeit at the expense of some image quality distortion due to the lossy nature of the encoding process. This trade-off should be considered for each task separately.

For applications requiring super-high performance in terms of raw image capture and encoding, we also have our 12-bit JPEG algorithm on CUDA, which is approximately more than ten times faster than the JPEG2000 encoder. For our test image, we've achieved a raw encoding frame rate of around 230 fps for 12-bit JPEG encoding on the Jetson Orin AGX with a jpg quality of 100 (meaning minor quality distortion). In this case, the compression ratio is around 15% lower than that of JPEG2000, but it is still very interesting and promising. With this algorithm, we also have the option of splitting the Bayer channels into four different planes and encoding them separately. This improves the compression ratio at the expense of some performance decrease.

There are also various applications which require high compression ratio for 12-bit RAW images from machine vision image sensors from SONY, Onsemi, Gpixel, etc. These use cases could be handled with our lossy RawZipper codec on CUDA, which is built for efficient and fast RAW Bayer compression with minimum visual image distortion.

We can offer an evaluation of our Jetson ISP software by request. Please fill in the form below and send it to us.

Other info from Fastvideo concerning JPEG2000 and Jetson

About the author

Fyodor Serzhenko, PhD, is the founder and CEO of Fastvideo. He earned his PhD at the Moscow Institute of Physics and Technology (MIPT) in 1993. Since 2009 he has led the development of Fastvideo’s GPU-accelerated image codecs and ISP modules, including the Fastvideo SDK. Connect on LinkedIn.

Why you can trust these results

Fastvideo has built GPU-accelerated image processing software since 2009. Our codecs comply with the standards they implement — JPEG (ITU-T T.81 / ISO IEC 10918) and JPEG2000 (ITU-T T.800 / ISO IEC 15444) — and our tools are open source on GitHub. Every performance figure we publish is reproducible: see our benchmark methodology, download the Fastvideo SDK benchmark report (PDF), and measure it on your own GPU.