Replacing JPEG 2000

No comments

ABSTRACT

This paper describes a drop-in replacement for the JPEG 2000 (J2K) block coder that offers exceptionally high throughput, with a small reduction in coding efficiency, while retaining all features of J2K except for quality scalability.

Throughputs on the order of 10x or more are achievable relative to J2K. We coin the term FBCOT (Fast Block Coding with Optimised Truncation) for the overall proposed algorithm. Truly reversible transcoding between J2K and FBCOT bit-streams is supported on a blockby-block basis, enabling systems in which the efficiency and scalability of J2K can be combined with the high throughput benefits of FBCOT.

Based on the outstanding performance of FBCOT, a new standardisation activity has been launched within the ISO/IEC working group known as JPEG.

INTRODUCTION

The JPEG 2000 (J2K) image compression standard provides high compression efficiency along with a unique set of features, enabling extraction of resolutions and spatial regions of interest directly from the coded representation. These features render it particularly valuable for interactive and cloud-based content distribution applications.

J2K is the basis of Digital Cinema and is used more widely within the video entertainment, broadcast and archival industries. We can expect to see J2K increasingly used in the future, since it offers highly efficient interactive access to content that is too large to consume or view all at once, including VR content.

This is enabled by the JPIP standard (J2K Part 9) for interactive transport of image/video content based on user defined windows of interest. For a comprehensive treatment of the JPEG 2000 standard and an introduction to JPIP, the reader is referred to Taubman and Marcellin (1) and Taubman and Prandolini (2).

The main obstacle to more widespread adoption of J2K in entertainment and broadcast sectors has been the relatively high computational complexity of its underlying block coding algorithm. This paper presents a drop-in replacement for the block coding algorithm that is capable of very high throughputs and energy efficient deployment in software and hardware.

Moreover, the new coding algorithm supports truly reversible transcoding to or from any J2K code-stream, on a block-by-block basis. This opens up a range of opportunities for new media compression solutions and systems. Early evidence based on the algorithm described in this paper has inspired the creation of a new activity within the ISO/IEC JTC1/SC29 technical working group WG1 (a.k.a. JPEG).

This activity, known as High Throughput JPEG 2000 (HT), is expected to culminate in 2018 with Part 15 of the JPEG 2000 family of standards after a competitive process.