Articles: Monitors
 

Bookmark and Share

(2) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 ]

Well, not exactly so. First of all, images are not of perfect quality themselves – being restricted by the capabilities of the camera, scanner or any other device you’ve made them with. The precision of darks is first of all determined by the noise in the camera’s CCD or CMOS array (there can be numerous reasons for the noise – photon shot noise, read noise, dark current of the matrix and so on). So, the signal-to-noise ratio, even of high-quality cameras with cooled matrices employed for scientific purposes (in astronomy, spectroscopy, microbiology and so on), is 60..65dB (to reach that, at least two-step Peltier-based cooling is employed, complemented with active air cooling of the heatsinks; the resulting temperature of the CCD array being from -10°C to -40°C). This corresponds to about 10-bit precision (1 bit = 6.2dB). Ordinary cameras, up to professional ones, provide a SNR of 40..50dB at best, which corresponds to a precision of 7-8 bits. So, what’s the sense of additional precision bits if the junior bit only contains the matrix’s noise at the standard 8-bit precision?

Moreover, gamma compensation reduces the color-reproduction precision by itself – due to truncation errors as well as during subsequent processing of the compensated images. These distortions are most noticeable in dark areas of the image, although gamma compensation is intended to render them with a higher precision than a linear display. Anyway, we’re not unlikely to see the industry to abandon gamma compensation in the near future. Too much equipment is designed to work with it.

But let us return to LCD monitors. I mentioned above that CRT monitors have an exponential display characteristic; with LCD ones it is close to S-shaped. In other words, to achieve the necessary exponential dependence on an LCD monitor, we need a correction table that translates the natural dependence into the required. Thus, the monitor’s color-reproduction properties will also depend on the profile the manufacturer flashes into the firmware. The manufacturers find a kind of compromise between calibration of each particular monitor (that’s unpractical with respect to the speed of the production line as calibration takes about a quarter of an hour or more) and a single-time calibration at the start of production of a new model (this is unacceptable from the point of view of quality – matrices from different batches may have different characteristics). But even a batch-specific calibration is no guarantee of the best result. The calibration curves for an Acer AL1715 monitor are shown in the picture below as an example (red, blue and green lines are the calculated curves for gamma=2.2; measured curves for the appropriate colors are given in black):

The graph reveals that the “native” S-shaped characteristic of the matrix is not fully compensated: the brightness swoops down in the middle of the range, and the lights, on the contrary, are displayed brighter than they should be. This deviation is rather small, and the home user won’t even notice it, but there can be worse cases:

This graph is drawn for an e-Yama 17JN1S monitor. As you see, the manufacturer overdid it – the colors are saturated earlier than the rightmost point of the graph. In practice this means that, for example, the color RGB:{224;224;224} will be displayed as pure white rather than light-gray. In other words, the monitor doesn’t distinguish between some light tones, reproducing them all as white color. A similar thing may happen to dark tones when the monitor displays a dark gray as black. Sometimes you can even witness a paradoxical situation when, for example, the color RGB:{5;5;5} is darker than the pure black RGB:{0;0;0}. The quality of the color reproduction setup also depends on the brightness and contrast settings.

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 ]

Discussion

Comments currently: 2
Discussion started: 01/01/14 09:32:13 PM
Latest comment: 01/02/14 08:41:35 PM

View comments

Add your Comment