How does HTJ2K compare to the venerable JPEG (Rec. ITU-T T.81 | ISO/IEC 10918), which remains in wide use for photographic applications?
JPEG 2000 has, in the past, failed to gain traction in applications covered by JPEG in part because of its lower relative encoding and decoding throughput performance. This is rectified by the HT block coder, which increases throughput by an order of magnitude over that of traditional JPEG 2000.
As illustrated in the following figure, single-threaded HTJ2K coding and decoding speeds (in orange) are in fact slightly higher compared to JPEG (in blue). In contrast, HTJ2K offers more opportunities for parallelization that JPEG and its advantage grows tremendously when multithreading is used (in red).
Furthermore, and as expected, HTJ2K (in orange) offers better image quality scores than JPEG (in blue) at a given bitrate, as illustrated below. SSIM is used here as a proxy for image quality.
The results were generated over the UHD ARRI Alexa Drums image sequence. Turbo JPEG was used to generate the JPEG results and Kakadu SDK for the HTJ2K results.
