Originally Posted by dmunsil
Quote:Originally Posted by John Mason
Yup, guess you can create all kind of tricks and measure false effective resolutions. I'd want to set the display accurately, as normally viewed, then measure the maximum effective horizontal resolution: a lens/reticle against a direct view screen--already known to fully resolve 1920X1080 from a test pattern--then comparing the observed reticle spacing(s) with a range of burst frequencies/resolutions to determine the maximum resolutions of the finest details.
OK, I understand now. And I think this could be an interesting and fruitful approach, but two caveats come to mind right off the bat:
- It's very difficult to tell the difference between an artifact and real content, especially on content transferred from film. Ringing can be obvious, or it can be much more subtle, and it can make a low-resolution area of the picture look like it has higher resolution. Film grain can add all kinds of fine details to the image, but it's obviously not adding any real content.
- It's going to be difficult to find a detail in a real movie with a hard edge that has a clear, unambiguous width. Most details in real-world content have been mildly smeared by a variety of anti-aliasing processes. You certainly won't find clean sine waves that you can just measure the peak-to-peak distance on. Say you look at a strand of hair under a loupe. It will almost certainly not be exactly one pixel wide. And it's impossible to measure exactly how much more or less than one pixel it is.
As Stacey implied, it's pretty much impossible to use the full resolution of any fixed-pixel display for an image without introducing ugly aliasing. Some of the details need to be smoothed, and that's usually done at least somewhat perceptually. There may be a default rolloff that tends to produce reasonably pleasing results (as there is in any digital camera), but if needed the filmmakers or transfer engineers can roll off a touch more to get rid of sparkles or line twitter or other issues caused by too much detail up near Nyquist.
Tried a few HD image measurements recently, using only a millimeter-marked ruler. Haven't acquired an optical loupe with a reticle yet, but while measuring I noticed a few instances where a magnifying loupe would have been useful.
You and Stacey were right about the difficulty measuring film detail resolutions. Tried a Blu-ray, The Tree of Life,(TOL)
initially, one of the top-rated 0-tier productions in the Blu-ray PQ thread here. First I measured the black-line width on screen for 5 B&W vertically lined multiburst frequencies on your test disc (for the highest static horizontal resolutions
). Frequency labels are not on the first test disc but appear, with TV-line nomenclature, on the 2nd-gen disc (see calibration forum). On my 65" Panny plasma, from a Sony PS3 machine, got these results:
Frequency Blk line width Res/Pic. Ht. 16X9 Res (x 1.78)
37.09 ~1 mm 1079 1920
18.55 ~2 mm 962 1712
12.36 ~2.5 mm 641 1140
9.27 ~4 mm 481 856
7.42 ~5 mm 385 685
Besides typical defocusing of backgrounds when characters were in the scene, forest shots--perhaps useful for fine details--often involved motion, blurring distant twigs etc.
In a medium-distance shot of a small store's facade, I measured a dark vertical line in the upper corner (~1 55 58 elapsed time) that's 2mm wide or ~1770 lines full 16X9 picture width resolution (962X1.84). (This 4k DI Blu-ray has a special 1.84:1 aspect ratio.) Most façade lines are still finer but barely visible inches or a few feet from the screen. A magnifying loupe and intermediate burst frequencies (>~19 MHz) would have helped. Earlier I posted sources
for loupes, reticles, etc.
A TOL close up of the wife's straggling hairs (~1 55 14) was easily measured. One hair width was also ~2mm (~1770 lines ). Her image occupied half the frame, and the lens magnification is a factor.
Also recently measured some fine details at a 1080i golf match on the Golf Channel (via NYC's FIOS from Verizon). In one medium-range zoomed shot, with a golf green occupying about 2/3 of the image, narrow marsh reeds in the top background aided realism. The full-sunlit side of each reed or grass blade measured about 2 mm or ~1712 lines for a 16X9 image (962X1.78), similar to the typical ~1700-line limiting horizontal effective resolution of 1080i HD broadcasts. -- John
Edit: Table resolutions above calculated from Don's formula earlier above
(F * 1920 / 37, where F=burst frequency).