Upcoming Media Formats Studied by JPEG
The Joint Photographic Experts Group (JPEG) is currently researching and developing several next-generation media specifications designed to address artificial intelligence, immersive displays, media authenticity, and extreme archival demands. While best known for the ubiquitous 1992 JPEG baseline standard, the committee's current exploratory and standardization activities focus on emerging paradigms through initiatives such as JPEG AI, JPEG DNA, JPEG Trust, and JPEG Pleno.
JPEG AI
JPEG AI represents the committee’s first effort to develop an end-to-end image coding system driven entirely by deep neural networks. Traditional codecs rely on manually engineered transforms, prediction, and entropy coding. In contrast, JPEG AI uses machine learning models to achieve superior rate-distortion performance compared to current state-of-the-art standards. Beyond optimizing images for human visual perception, JPEG AI is engineered to facilitate machine-task efficiency, enabling algorithms to perform classification, segmentation, and object detection directly on compressed representations without requiring full image reconstruction.
JPEG DNA
JPEG DNA explores the use of synthesized biological DNA for digital media storage. Faced with exponential growth in media creation and the physical limitations of magnetic tape and optical media, the committee is studying methods to transcribe digital imagery into sequences of the four DNA nucleotides: Adenine (A), Cytosine (C), Guanine (G), and Thymine (T). This initiative aims to define compression and error-correction architectures tailored to biochemical synthesis, storage, and sequencing, offering data retention spans measured in centuries with unprecedented storage density.
JPEG Trust
The rise of generative synthetic media and sophisticated editing tools prompted the creation of JPEG Trust. This standard establishes an interoperable framework to establish authenticity, provenance, and intellectual property rights for digital media. JPEG Trust does not evaluate whether an image is "real" or "fake" algorithmically; instead, it defines a structured architecture for cryptographically secure metadata that tracks the entire lifecycle of an asset—from original sensor capture or algorithmic generation through every subsequent edit, compression, and publication step.
JPEG Pleno
JPEG Pleno provides a unified framework for capturing, representing, and rendering multimodal, high-dimensional visual information. Unlike traditional flat-raster images, JPEG Pleno addresses plenoptic imaging modalities that capture the full physical behavior of light within a scene. The primary modalities under this initiative include:
- Light Fields: Capturing the intensity and direction of light rays to allow post-capture refocusing and perspective shifting.
- Point Clouds: Representing three-dimensional geometric data alongside color attributes for volumetric scenes.
- Holography: Encoding the phase and amplitude of wavefronts to display true optical depth without specialized eyewear.
Emerging Explorations
The committee also maintains active working groups evaluating real-world interactions with media assets. Explorations include JPEG NFT, which examines metadata consistency and asset integrity across blockchain ecosystems, and assessments of neural rendering formats (such as NeRFs and 3D Gaussian splatting) to evaluate their potential integration into future standardized interchange formats.