Full Description
Scope
This standard defines a set of tools for efficient image coding, including tools for encoding, for decoding, and for encapsulation. All or some of the tools may be based on trained neural networks, and may perform block partitioning, prediction, transform, quantization, entropy coding, filtering, etc. Purpose
This standard provides efficient, neural network-based coding tools for compression, decompression, and packaging of image data, which significantly improve the compression efficiency compared to IEEE Std 1857.4 (intrapicture coding) and IEEE Std 1857.10 (intrapicture coding) under comparable settings, and facilitate the compression and decompression on top of neural network-oriented computing infrastructures like neural network processing units (NPUs). The target applications and services include but are not limited to Internet images, user-generated images, and other image-enabled applications and services such as digital image storage and communications. Abstract
New IEEE Standard - Active - Draft.This standard defines a set of tools for efficient image coding, including tools for encoding, for decoding, and for encapsulation. Some of the tools are based on trained neural networks. These tools are designed to perform block partitioning, prediction, transform, quantization, entropy coding, filtering, etc., respectively.
