The white paper describes a drop-in replacement for the original JPEG 2000 block coding algorithm, along with its implications for the whole JPEG 2000 family of standards and for a variety of image and video compression, archival and rendering tasks. In this document, we refer to the original block coding algorithm, defined in JPEG 2000 Part-1 as the “J2K-1” algorithm, while the new block coder is known as the “HT” (High Throughput) algorithm.

Terminology and Standards:

HT = “High Throughput”

HTJ2K code-stream = a coded image representation conforming to ITU-T Rec T.814 | IS 15444-15, read as “High Throughput JPEG 2000, which is based on the representation in JPEG 2000 Part-1 and compatible with the features of JPEG 2000 Part-2. HTJ2K code-streams can contain J2K-1 code-blocks, in addition to HT code-blocks.

JPH = an image file format described in ITU-T Rec T.814 | IS 15444-15, that is based on the JP2 format in JPEG 2000 Part-1, with enhanced colour space and colour association capabilities, embedding a single HTJ2K code-stream.

FBCOT = “Fast Block Coding with Optimized Truncation” is an algorithm that can be used by HTJ2K encoders to achieve a target compressed size without iterative encoding, similar to the well-known EBCOT algorithm that is used with J2K-1. FBCOT works with the HT block coder and may include additional complexity management.

“Importantly, the HT block coding algorithm allows completely lossless transcoding to/from code-streams that have been generated using the J2K-1 algorithm.”

At quality levels of interest to most applications, the HT block coding algorithm is an order of magnitude faster than J2K-1 in a carefully optimized CPU-based implementation, as determined by a careful evaluation during the development of the HTJ2K standard by the ISO/IEC working group known as JPEG (formally ISO/IEC JTC1/SC29/WG1). See details in the white paper.

As shown later in this document, the entire end-to-end encoding or decoding system can be almost 10 times faster for an HTJ2K code-stream than one based on the J2K-1 algorithm, at typical compressed bit-rates employed in professional applications, with much larger speedups on the order of 30x for lossless compression or decompression.

As a drop-in replacement for J2K-1, the HT block coding algorithm can be used in conjunction with pretty much any of the technologies described in the entire JPEG 2000 family of standards, to dramatically increase throughput, while retaining most existing features.

Importantly, the HT block coding algorithm allows completely lossless transcoding to/from code-streams that have been generated using the J2K-1 algorithm.

The price paid for the dramatic reduction in complexity associated with the HT block coding algorithm is a modest reduction in compression efficiency (typically 6% to 10%), plus substantial elimination of the quality scalability feature of JPEG 2000.