View Full Version : what is "sharpness" and why does it matter?


jon_vogel
04-12-07, 11:09 AM
I've been puzzled while reading about the "sharpness" of one 1080p projector over another.

First I'll state a few of my assumptions:

1) I assume people are _not_ talking about artificial edge enhancement of the picture (TV sharpness control).
2) I assume people _are_ talking about the ability of the chip panel and optics to put a fully resolved crisp square pixel on the screen.

So on a "sharp" projector I can walk up to the screen and see a nicely defined square. On a not-so-sharp projector the pixel may have blurred edges or not look precisely like a square.

Are these assumptions correct?

If so, I don't see why this matters for video/movies. If I can see the pixel structure from my viewing position, it takes me out of the movie's reality. It makes the screen seem less like a (slightly blurry) window to another world and more like a digital computer graphics business presentation.

In fact, on my current 720p DLP projector I slightly defocus the lens to avoid SDE and defined pixels.

Now, I'm all for better optics to reduce light scatter and improve ANSI contrast, etc. But I don't get why we need to resolve information smaller than one pixel.

Am I missing something?

MrWigggles
04-12-07, 11:46 AM
For someone who only has 21 posts you do have a good understanding of sharpness.

The definition of sharpness is debatable. (Does sharpness mean sharp edges, good MTF, resolution, pixelization, etc.) What I would like to talk about is fidelity. Faithfully reproducing the content.

Video is a little different than computer graphics because there is an assumption that you are not trying to create little squares (i.e. pixels) on the screen. With video, it is assumed that there is supposed to be some level of smoothness.

Video is a two-dimensional waveform and is similar in principal to audio when it comes to analysis. Pixels actually create harmonics. A pixel wide series of black and white lines is a squarewave. This squarewave is what makes images "look digital" when viewed from up close; your eye is resolving the fundamental and the harmonics. For computer text programs that is fine, but for video it is not. Visually, an alternating one pixel wide set of lines should be a sinewave and thus smooth with no harmonics.

Well, now that we know what we want, how do we get there? If you have a 1920 X 1080 projector and you feed it an image with 1920 horizontal lines, you obviously can't solve the problem digitally (there are no in between pixels) so you either have to solve it through panel technology or defocused optics (defocused optics could just be bad eye-sight of the viewer). For DLP and LCoS defocusing the optics is aways an option, but LCoS has an advantage that there is some pixel bleeding that happens in the panel itself.

However the defocusing is done, the goal is to remove the harmonics of the squarewave. Ironically the limitations of most optics has made the defocusing built-in; the best focus you can achieve on 1080p projector is probably defocused enough already.

Some LCD and LCoS projectors have so much defocusing in the optics that they will add some pre-emphasis to the higher frequencies (i.e. digital sharpening) which the optics will then attenuate.

I could go on and on. I will just add that for most 1080p projector out now, overly sharp images aren't a problem. The problem is often a lack of sharpness. If you can't see the pixel structure from up close than chances are projector is not overly sharp and likely isn't sharp enough.

I wish I could give you a more quantitative answer than what I just wrote, but it will require some detailed math, illustrations, and research into SMPTE standards that I can't quite quote off the top of my head.

-Mr. Wigggles

R Harkness
04-12-07, 12:47 PM
Or, to approach the question another way in a nutshell:

We want sharpness without visible pixel/screen structure. One of the benefits of higher res 1080p displays is the ability to have both, because you can have great sharpness, while the pixels become small enough that you don't see the screen structure. (Unlike 720p projectors where the pixel structure tends to become visible much sooner as you move toward the image).

When it comes to the details of any particular image, you have to make the distinction between sharpness in terms of "resolution" and sharpness in terms of "apparent sharpness." As you know , tricks like edge enhancement (adding white outlines around images) can increase apparent sharpness without increasing actual resolution (that is, fine detail is not actually increased, details are just more "brought to your attention" by giving them a bright halo).

Resolution is not the same as sharpness, but resolution (with the right source material) tends to result in finer definition of detail and hence better image sharpness. That's one benefit. Also, all other things being equal, an increase in image sharpness will increase the sense of image realism, including increasing dimensionality.

TF Ghost
04-12-07, 01:01 PM
Excellent thread! Is there anyway to make explanations of fundamentals, such as this, sticky's or maybe create a new forum for noobs (like me) looking for info like this? (Or maybe there is already and I've missed it in my noobdom)

Craig Peer
04-12-07, 02:25 PM
We want sharpness without visible pixel/screen structure. One of the benefits of higher res 1080p displays is the ability to have both, because you can have great sharpness, while the pixels become small enough that you don't see the screen structure. (Unlike 720p projectors where the pixel structure tends to become visible much sooner as you move toward the image).

Absolutely correct !! I found when I went to demo a SIM C3X that it had a sharp detailed picture ( that I loved ), but also sharp detailed screen door compared with my H79 from about the same viewing distance ( that I hated ). My dVision 1080p by contrast has more detail, is just as sharp, but no screen door at even sea sickness inducing close distances from the screen!

MrWigggles
04-12-07, 02:33 PM
In a kind of related side-topic, flat panel displays will always look pixelated. There is no way to defocus the image. So while I think 1920 X 1080 is the final frontier for Front Projectors (with today's content), flat panels still need more resolution.

Flat panels "need" a 3840 X 2160 panel resolution to reproduce 1920 X 1080 images really really well. They need transition pixels to smooth the input image.

Viva La FP!!

-Mr. Wigggles

TF Ghost
04-12-07, 02:41 PM
In a kind of related side-topic, flat panel displays will always look pixelated. There is no way to defocus the image. So while I think 1920 X 1080 is the final frontier for Front Projectors (with today's content), flat panels still need more resolution.

Flat panels "need" a 3840 X 2160 panel resolution to reproduce 1920 X 1080 images really really well. They need transition pixels to smooth the input image.

Viva La FP!!

-Mr. Wigggles

Matt?

Oiler
04-12-07, 02:50 PM
While I would be the last person to bemoan the tremendous price drops
we have seen lately, I cannot help but noticing that while some features
have improved dramatically, others are starting to disappear. Most of these
seem to be related to the optics directly or indirectly. Quality lenses for example are not cheap and I do not get the sense that this will change in the near future. As such this appears to create a cost barrier with respect to certain aspects of pq.

R Harkness
04-12-07, 03:19 PM
In a kind of related side-topic, flat panel displays will always look pixelated. There is no way to defocus the image. So while I think 1920 X 1080 is the final frontier for Front Projectors (with today's content), flat panels still need more resolution.

Flat panels "need" a 3840 X 2160 panel resolution to reproduce 1920 X 1080 images really really well. They need transition pixels to smooth the input image.

Viva La FP!!

-Mr. Wigggles

Interesting, I don't find that to be the case. I've demoed the latest 1080p plasmas (Panasonic/Pioneer) more times than is healthy for me :) and I find the pixel structure invisible from any distance I'm likely to view the panel. (E.g. from beyond 5 feet I no longer can see pixel structure on the Panasonic 65 1080p model, and I would never be watching it from that close. At up to 8.5 feet, as implied by acuity charts, I would still perceive the added resolution from 1080p while not necessarily perceiving pixel structure).

MrWigggles
04-12-07, 03:34 PM
Rich,

Have you seen the QuadHD displays? There is a visible difference even from 5 feet on the prototypes shown at CES. In fairness, I don't know what the content was on the QuadHD monitors, but it was obviously more detailed and pixelated than the 1080p even though the 1080p monitors didn't look pixelated.

When I took my glasses off, everything looked the same.

-Mr. Wigggles

scaesare
04-13-07, 09:32 AM
I'd like to add that, IMO, it's this issue that leads some to claim that DLP is sharper than other technologies... often with the implication that it's therefore more true to the source (i.e. it's not "softening" the source).

While DLP may indeed be sharper, I believe it's often a "false sharpness" due to the high-frequency harmonics that the mirror edges produce, and to which Wigggles refers.

While it may appear sharper to LCoS/SXRD in comparison, the true comparison should be real life. It often looks artificial from that standard. Certainly this effect is lessened on 1080p devices over 720p ones. This may be another place where greater-than-1080p would be useful for "transition" pixels.