Και επειδή ξέρω ποσο γουστάρεις το θέμα, ιδού μια εξαιρετική ανάλυση από τον Greg Rogers που εξηγεί τις διαφορες Ansi contrast, on-off contrast αλλα και την επιρροή του gamma στο contrast:
Specs
I believe the original question inferred in the title of this thread is whether or not you should ignore contrast ratio specs. Since contrast ratio specs are established by the projector manufacturers, who almost never publish sufficient controlling operating conditions for their contrast ratio specs (which would have to be the same for all manufacturers, and would ideally be applicable to well calibrated home theaters), I would never rely on the published contrast ratio specs when comparing projectors.
Measurements
I think the thread then evolved into a more important, and more basic question. Should you ignore the two traditional contrast ratio measurements, full-field (often called on-off) contrast ratio and an ANSI-like (my m-ANSI for instance) contrast ratio, when they are competently made under controlled and reproducible conditions? I believe you should not ignore those important measurements. All other things being equal, bigger is always better for those two contrast ratio metrics, for both fixed and dynamic iris projectors.
These two measurements provide information about two very different physical phenomena that ultimately interact in extremely complex ways to affect our perception of intra-image contrast in real images. The full-field contrast ratio is really an indirect method of specifying the minimum black level relative to the reference white level, and therefore the minimum black level relative to brightness of all other signal levels (i.e. the contrast of all other signal levels to black --- gamma also plays a essential role in that contrast). Consequently, the full-field contrast ratio (along with gamma) determines the contrast between dim objects (shadow detail) in a dark image (i.e. in the absence of bright objects). So although we characterize full-field contrast ratio with large numbers (5,000:1 - 30,000:1) it really affects the much smaller contrast ratios (10:1, 20:1, 30:1) between dim objects and dim objects to the dark background that we perceive as major differences in image quality.
The ANSI-like contrast measurements with checkerboards are a way of characterizing a projector's light scatter from bright objects in the image to other parts of the image. We use a standard 4x4 checkerboard to make the measurement and that directly influences the magnitude of the ANSI (or my m-ANSI) contrast ratio. If we use a 8x8 checkerboard pattern we get a much smaller contrast ratio value because the measuring points within the dark checkerboard blocks are much closer to the surrounding bright blocks. That simple experiment and others we can conduct, shows that the contrast ratio from light scattering depends on the proximity and distribution of bright image areas relative to the location of the contrast measurement. Hence, the ANSI contrast measurements are simply a figure of merit, where bigger is better (less light is scattered from bright objects), but we can't directly use the measured value to compute intra-image contrast in real images that have a different distribution of bright and dark areas.
Keep in mind that each image of approximately 2 million pixels may have millions of contrast ratio relationships, although only a small subset (perhaps dozens) of those contrast ratios may dictate our perception of a given image at any particular instance. However, simple analysis (confirmed by trivial observation) shows us that increasing the full-field contrast ratio primarily improves intra-image contrast performance in dark images, while increasing the ANSI contrast primarily improves the intra-image contrast performance in images that have very bright regions (particularly when very dark regions are present in the same image). The most relevant effects of ANSI contrast can be complex, and non-intuitive. The light scatter (characterized by ANSI contrast measurements) that reduces the contrast between two dim objects (veiling effects) is often more important to the perception of image quality than the much larger contrast ratios between a bright object and dark object in the same image.
Gamma
What shouldn't be forgotten is the huge affect of gamma on intra-image contrast. For instance, suppose that we have an image where an area of 20% luma amplitude is surrounded by an area of 80% luma amplitude. (Perhaps the dark area is a small partially shaded park block in the middle of a bunch of tall sun-lit buildings.) And suppose that the size and distribution of the bright area around the dark area is very much like a 4x4 checkerboard, so the light scattering effects are virtually the same as those found in an m-ANSI measurement. (That is contrived for simplicity, but there will examples in complex bright images where the scatter ratio will happen to be equal to the m-ANSI measurement, but there is nothing special about using that exact value in this example.) Then suppose we have an RS2 projector with a full-field contrast ratio of 30,000:1 and an m-ANSI contrast ratio of about 300:1. So for this example, about 1/300 (black level is negligible) of the brightness of the buildings (which is produced by an 80% luma amplitude) is scattered over the center of the park block (which would otherwise have a luminance determined by its 20% luma amplitude). Under these conditions, the contrast ratio of the surrounding buildings to the center of the park block would be 19.7:1 if the projector gamma is set to 2.2. Now suppose that we magically make the m-ANSI contrast ratio of the RS2 infinite (there is zero light scattered from the buildings onto the park block). The contrast ratio would change to 21.1:1. That is an increase of only 7%. But what happens if instead of having an infinite m-ANSI CR we just change the RS2 gamma from 2.2 to 2.4? The contrast ratio would increase to 25.5:1, which is an increase of 29%. So for this particular example, the increase in gamma (which we can do because of the high full-field contrast ratio of the RS2) is much more important to the intra-image contrast than an increase in the m-ANSI contrast ratio.
Now suppose that in the same image there is a window in one of the buildings and the window also has a 20% luma amplitude. But because the window is much smaller the light scattering contrast ratio from the surrounding building to its window is 75:1 (rather than 300:1 in the park). In this case, changing the gamma from 2.2 to 2.4 increases the contrast between the window and building by 23%, but increasing the m-ANSI contrast ratio to infinite increases the contrast by 28%. However, just increasing the m-ANSI to 900:1 (3x increase - a much more reasonable target) increases the window contrast by 17%.
So within this example of a fairly bright image, an increase of gamma from 2.2 to 2.4 is very significant, but doubling the m-ANSI contrast approaches similar effects in areas where the light scattering is strongest, but would have little effect in areas where the light scattering is weaker. If we looked at examples where light was scattered from bright objects to shadow detail, the effects from m-ANSI contrast ratios dominate the effect of gamma changes, and in fact increasing gamma can even become mathematically counter-productive in the right circumstances. This is already a very long post, so I won't go through one of those examples. But in an image of a real city there may of course be dozens (or even hundreds) of additional contrast relationships in the image that are all affected by full-field contrast, m-ANSI contrast (which is affected by the brightness, size and distribution of image features), and gamma in determining the perceived intra-image contrast.
__________________
Greg Rogers
AccuPel
Widescreen Review
