The original JPEG algorithm in ITU-T Rec. T.81 | ISO/IEC 10918-1 (“JPEG-1”)
contains two entropy coders: Huffman coding and an arithmetic coder called
the "QM-coder". The Huffman coding method is required in the T.81 baseline.
For all other encoding processes (extended sequential, progressive,
hierarchical, and lossless), the use of all defined entropy coders is
T.851 (“ITU-T T.81 (JPEG-1)-based still-image coding using an alternative
arithmetic coder”) defines a still-image coding method based on ITU-T Rec.
T.81 (JPEG-1) which makes use of an alternative arithmetic encoder called
the "Q15-coder", and defines an "alternative baseline" compared to ITU-T
Rec. T.81. Support for Huffman coding is not required in T.851.
It should be noted that images encoded according to this Recommendation can
be losslessly transcoded to or from any of the entropy coding methods of
ITU-T Rec. T.81, including the baseline Huffman coding of ITU-T Rec. T.81.
Interoperability with the Huffman-based baseline of T.81 is not required by
this Recommendation, but might be required by an application incorporating
T.851. An application specification might require a dual-mode entropy codec,
or a server-based transcoding facility.
Since the Q15-coder resolves the carry in the decoder instead of in the
encoder, it has a low latency compared to the QM-coder arithmetic coder
incorporated in T.81, which has a potential latency of the rest of the
entire entropy-coded segment.
This Recommendation also specifies a DCT with input precision of 16 bits,
which is greater than the precision of the T.81 (JPEG-1) DCT.