pSQ - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 61

ponding step size either Δ s or Δ n ; thus, I ρ is associated with a specific interval on the real line.
Then, the perceptually quantized coefficients Q , from a known viewing distance d , are calcu-
lated as follows:
(2)
Unlike the classical techniques of visual frequency weighting (VFW) on JPEG2000, which
apply one CSF weight per sub-band [ 1 , Annex J.8], perceptual quantization using CBPF(pSQ)
applies one CSF weight per coefficient over all wavelet planes ω s , o . In this section, we only ex-
plain forward perceptual quantization using CBPF (F-pSQ). Thus, Equation (2) introduces per-
ceptual criteria of the perceptual images (1) to each quantized coefficient of the dead-zone scal-
ar quantizer [ 1 , Annex J.8]. A normalized quantization step size Δ = 1/128 is used, namely, the
range between the minimal and maximal values at I ρ is divided into 128 intervals. Finally, the
perceptually quantized coefficients are entropy coded, before forming the output code stream
for bitstream.
The perceptual quantizer F-pSQ in JPEG2000 is tested on all the color images of the Mis-
cellaneous volume of the University of Southern California Image Data Base [ 6 ] . The data sets
are eight 256 × 256 pixel images and eight 512 × 512 pixel images, but only visual results of
the well-known images Lena , F-16 , and Baboon are depicted, which are 24-bit color images and
512 × 512 of resolution. The CBPF model is performed for a 19 in. monitor with 1280 pixels of
horizontal resolution at 50 cm of viewing distance. The software used to obtain a JPEG2000
compression for the experiment is JJ2000 [ 7 ].
Figure 1(a) shows the assessment results of the average performance of color image com-
pression for each bit-plane using a dead-zone uniform scalar quantizer (SQ, function with
heavy dots), and it also depicts the results obtained when applying F-pSQ (function with
heavy stars).
FIGURE 1 (a) JPEG2000 Compression ratio (bpp) as a function of Bit-plane. Function with
heavy dots shows JPEG2000 only quantized by the dead-zone uniform scalar manner. While
function with heavy stars shows JPEG2000 perceptually prequantized by F-pSQ. (b) The bit-
rate decrease by each Bit-plane after applying F-pSQ on the JPEG2000 compression.
Using CBPF as a method of forward quantization, achieves beter compression ratios than
SQ with the same threshold, obtaining beter results at the highest bit-planes, since CBPF re-