Overview of an HTJ2K Encoding System

Figure 1 illustrates the elements of an HTJ2K encoder. HTJ2K substantially preserve the existing architecture and code-stream syntax of JPEG 2000. Imagery is first subjected to any re-quired multi-component transforms and/or non-linear point transforms, as allowed by Part-1 or Part-2 of JPEG 2000, after which transformed image components are processed by a reversible or irreversible Discrete Wavelet Transform (DWT), which decomposes each component into a hierarchy of detail sub-bands and one base (LL) sub-band.

Figure 1: HTJ2K encoding system based on the FBCOT (Fast Block Coding with Optimized Truncation) paradigm.

Apart from the block coding algorithm itself, the HT block coder has no substantial impact on other JPEG 2000 code-stream constructs or their interpretation.  Precincts, packets, tiles, transforms, and all other JPEG 2000 elements remain substantially unchanged, except in the so-called MIXED mode, where J2K-1 and HT block coding algorithms can both be used within a single tile-component.

In general, the construction of JPEG 2000 packets depends upon a set of codeword segments that are produced by the block coder. In its default mode, the J2K-1 block coder produces only one codeword segment, containing all coding passes, but there are other modes in which the coding passes are partitioned into multiple codeword segments. The significant property of a JPEG 2000 codeword segment is that the total number of bytes within the segment is explicitly encoded within packet headers.