X-bit labs

EPSON Perfection 2480/2580 PHOTO: Fall Offsprings in the “Perfect” Family (page 5)

Category: Other

by Sergey Samarin

[ 09/27/2004 | 09:01 AM ]

Image Noises

I told you about the origin of digital noise in my previous articles. The signal-to-noise ratio allows evaluating the tolerance of the scanner’s electronics to various interferences. The noise can be random or correlated. Let’s deal with each type independently.

Random Noise

The graphs below show the dependence of the random noise value on the reflection power of the grayscale sectors of the KODAK IT8.7/2 Q-60R2 target.

The dependence of the random noise value
on the reflection power for the Perfection 2480…

…and for the Perfection 2580.

The signal-to-noise ratio, which is the ratio of the median of a grayscale sector to the deviation, should be regarded as “bigger is better”. That is, the bigger the SNR, the higher the scanner’s noise tolerance is. The delta SNR parameter is the total of the measurements.

Bigger delta SNR values indicate that the scanner
is well protected against interference

As you see, although the scanners have the same stuffing, the experimental data are different. I’m inclined to write this discrepancy off to the irregularity of the scanners’ lamps, though – that’s natural with mercury-based fluorescent light sources.

Correlated Noise

Correlated forms of noise are the most annoying – they show themselves as image “patterns” (usually horizontal or vertical stripes). The predominance of such noise greatly reduces the signal-to-noise coefficient, which is calculated as the ratio of the median to the deviation. The calculated coefficients are put into the next diagram. Bigger values are better: