The relatively new mathematical field of
wavelet transforms achieved a major success when the Federal Bureau of Investigation
decided to adopt a wavelet-based image coding algorithm, referred to as the wavelet/scalar
quantization (WSQ) standard, which is able to produce archival-quality images at
compression ratios of around 20:1. Besides a very specific wavelet subband decomposition
structure, the FBI standard is based on a highly optimized uniform scalar quantization
strategy derived as the solution to a nonlinear optimization problem of a high-rate
distortion model subject to a linear constraint on the overall bit rate and convex
nonnegativity constraints on the individual bit rates used to encode the wavelet subbands.
This contribution intends to give strong evidence that the FBI standard should be
redesigned in the sense that the entropy coder in use should be replaced by a lossless
zerotree coder. It is essential to notice that our approach differs fundamentally from
other applications involving zerotree coders because we use zerotree coding in a lossless
mode while practically all currently existing zerotree applications do not separate lossy
embedded quantization and lossless entropy coding. To proof that this separation pays off,
our approach is compared to the JPEG standard, the FBI standard and the conventional
embedded zerotree coder. Results given clearly suggest that our approach outperforms the
alternatives mentioned and should thus offer great potential for applications involving
image transmission and archival.