Setting contrast with Spears and Munsil disc - Page 7 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #181 of 484 Old 04-17-2010, 12:37 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:


Is this the pixel you mean?

No, I’m using color cop that references the upper left corner as the origin. I find several 255s in the blown out highlights. These are already blown out or fully saturated channels, not so much indicative of what would be blown out, but what is already blown out details. IMO.

I'll try to do an A/B with/without for perceptual value if I find the time. It would help to have the native PNG as I can't do much operations at all in jpeg without levels shifting around.

Much like with AlluringReality caps. I think A/B side-by-side helps here...

Quote:
Originally Posted by dlarsen View Post

And in an A/B manner for a perceptible delta test...


Dave
dlarsen is offline  
Sponsored Links
Advertisement
 
post #182 of 484 Old 04-17-2010, 12:52 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by dlarsen View Post

No, I'm using color cop that references the upper left corner as the origin. I find several 255s in the blown out highlights. These are alrady blown out or fully saturated channels, not so much indicative of what would be blown out, but what is already blown out details. IMO.

Dave

1130,656 ( bottom left origin) I do get 251 249 255 so like you said there are some suspect pixels in there.

However lets not drift too far off track guys or over analyse something. Samples from commercial BD and DVD is what we should spend the effort on.

digital film janitor
Mr.D is offline  
post #183 of 484 Old 04-17-2010, 12:57 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by dlarsen View Post


Much like with AlluringReality caps. I think A/B side-by-side helps here...



Dave

To be honest I did immediately think I saw a bit of a a difference on the side of his glasses in the top example and it does seem to excurse above 235 in this area mainly in red when measured.

Also the door handles on the cab.

And the hots on the wall in the last one ...if anything thats most obvious if I compare.

However the actual difference is subtle on a frame by frame basis . I do wonder how subtle it will look on a large screen (I'm currently viewing this on my 15" powerbook.

digital film janitor
Mr.D is offline  
post #184 of 484 Old 04-17-2010, 01:13 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by Mr.D View Post

To be honest I did immediately think I saw a bit of a a difference on the side of his glasses in the top example and it does seem to excurse above 235 in this area mainly in red when measured.

Also the door handles on the cab.

Fair enough. That's the value in A/B side-by-side to me. To my perceptions, they are pretty much indistinguishable. The blown out highlights are blown out highlights in both cases. About equally. It wouldn't matter if I cal'd for StudioRGB or sRGB, they'd still be blown-out highlights.

I would and do notice the extra brightness and contrast over the valid signal range if I abandon the notion of valid BTB/WTW image information in all mainstream consumer material however, as well as opening up the option for the other possibilities I've noted before.

Quote:
Originally Posted by Mr.D View Post

However the actual difference is subtle on a frame by frame basis…

Are you suggesting that there is a temporal component to this? Sorry, I don’t see a basis for this if so. Motion pictures are frame by frame and if it’s not substantially manifest in full frame static, then I don’t see how any temporal effects bear into the discussion.

If anything, I think static side-by-side should amplify the perception and discrimination by the HVS. Static side-by-side affords the opportunity to go back-and-forth and observe for a long time. I believe it's widely accepted that the HVS perception and discrimination is drastically reduced for any comparisons in isolation (no B) or for short time periods. At 60Hz, a frame is only present for ~17mS.

Dave
dlarsen is offline  
post #185 of 484 Old 04-17-2010, 03:40 PM
AVS Special Member
 
alluringreality's Avatar
 
Join Date: Jul 2006
Posts: 3,143
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 17 Post(s)
Liked: 34
My prior post with images from Almost Famous was a mistake. When I started looking at VC-1 video, I realized the 16-235 output setting hadn't worked. The quick way to verify the prior images were incorrect is that the border is at 0 instead of 16 like in the following.



Since Wedding Crashers was released on Blu-ray since the first time I looked at above white, I decided to go back to the scene that I previously posted. With the higher quality encode I'm not seeing the reflections on the glasses go above white like with the DVD. The Girl Next Door and Collateral, the other movies I previously posted from DVD, have also been released on Blu-ray. If I had to make a bet I would go for gray only going to reference white, because so far I have not found anything clearly above white on the Mpeg4 or VC-1 Blu-rays I've looked at.


LL
LL
alluringreality is offline  
post #186 of 484 Old 04-17-2010, 03:46 PM
 
ChrisWiggles's Avatar
 
Join Date: Nov 2002
Location: Seattle
Posts: 20,730
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
Not everything is more easily visible with static images. In my opinion I agree that clipping is generally more easily visible with static images, but banding for instance is often much MUCH more easy to see on moving images.

But I could conceive situations where an image brightens and you can see clipping unfold or detail become lost as this happens dynamically which you could see more easily as it unfolded across several images. You can see this not so often with clipping, but with shadow detail that appears or disappears particularly with fades. In any case, not all video artifacts are easier to see on a single static frame, so we should keep in mind that these comparisons are not identical to watching the content in real-time. In many cases viewing static frames can exaggerate our ability to discern artifacts, but in some cases it obscures those differences. We can't make universal statements that we will always see more artifacts with still-frames, because that certainly is not true. We should simply acknowledge that these viewing examples are informative, but take care to understand that they can also be misleading because in some cases they obscure artifacts, and in others they exaggerate them.

But we encounter similar kinds of differences even when viewing the content real-time, certain systems and certain viewing ratios may hide or exaggerate certain artifacts, so we should always be careful because the visibility or lack of visibility of artifacts is always affected by the viewing system and the viewing environment, so there are always limits to what we can glean from particular observations. Viewing observations are very informative, but never absolutely definitive.
ChrisWiggles is offline  
post #187 of 484 Old 04-17-2010, 03:54 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by Mr D View Post

1130,656 ( bottom left origin) I do get 251 249 255 so like you said there are some suspect pixels in there.

Agreed. Suspect. In addition to the 255s (that shouldn’t be there for proper StudioRGB), I measure lots of 250+s in the R and G channels. This seems to be in discrepancy to what is shown on the waveforms for this image.



The waveform plots don’t seem to show anything above about 237/238 (a guess, but nowhere near 250) for the R/G channels yet I measure lots of these in the image.

Dave
dlarsen is offline  
post #188 of 484 Old 04-17-2010, 04:21 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Okay so far so good.

I think we should guard against going off track though.

Really what I'd like to see is people posting some frames and saying what the situation is above 235.

I'm suspecting 240 is maybe the absolute max we need to be clipping at... rather than 254. However I have been finding this evening I'm still not so cool with how my displays look clipping any lower than 254 ( ie. not clipping).

So keep posting grabs so we can get a reasonable sample size: do try to quantify the accuracy of your imagery. Ideally we want it unmessed with but use test patterns to confirm your frame grab generation as at least being meaningful above 235.

Don't get caught up in trying to overly nitpick someone elses grabs... dave! If we get 100 samples and 80 of them don't have level above 235 then that would be pretty good evidence to base a decision on.

10 people do 10 grabs ( verify your levels to some degree please) and we should be able to say something meaningful.

digital film janitor
Mr.D is offline  
post #189 of 484 Old 04-17-2010, 04:39 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Fair enough,

I will note that sliding from 254 to 240 is a pretty significant backslide. If you slide any lower than 240, pretty soon you'll be at 235!

No caps from me today. I've done enough over the years and posted them here A/B. I've had more than 10 deleted. I've settled (never really left) pretty much where AlluringReality has, and that's just about where the reference dots are place in DVE and where the reference lines are drawn on Chris' app, and where the reference cursors are placed on the Leader waveform monitor, and the Tektronix analyzer, and....

Dave
dlarsen is offline  
post #190 of 484 Old 04-17-2010, 04:46 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by dlarsen View Post

Fair enough,

I will note that sliding from 254 to 240 is a pretty significant backslide. If you slide any lower than 240, pretty soon you'll be at 235!

Dave

Yes but if people were a bit more open about finding stuff out without getting argumentative we wouldn't still be kicking these topics about.

So lets just find out what the actual situation is. That way we can actually tell people something useful rather than just recommend a side to take.

This is 2010 guys surely we can figure out where white actually is.

digital film janitor
Mr.D is offline  
post #191 of 484 Old 04-17-2010, 05:21 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by dlarsen View Post



Are you suggesting that there is a temporal component to this? Sorry, I don't see a basis for this if so.

Dave

No I'm not suggesting that at all. You have one version of the frame "by" (next to) the clamped version of the frame in your examples.

You are coming across as if you are arguing for arguments sake Dave. Please stop it I am interested in getting this topic put to bed once and for all and the only thing that will do that is analysing footage in a meaningful way. Not arguing over semantics.

digital film janitor
Mr.D is offline  
post #192 of 484 Old 04-17-2010, 06:58 PM
AVS Addicted Member
 
millerwill's Avatar
 
Join Date: Apr 2004
Location: Berkeley, CA
Posts: 11,381
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 21 Post(s)
Liked: 29
Quote:
Originally Posted by Mr.D View Post

... I am interested in getting this topic put to bed once and for all and the only thing that will do that is analysing footage in a meaningful way. Not arguing over semantics.

Thanks, Mr. D. It's beginning to sound like it may be recommended to set Cst high enough so that the M&S white bars are washed out (clipped) down to ~ 240 to 245. I'll stay tuned.
millerwill is offline  
post #193 of 484 Old 04-18-2010, 02:09 AM
AVS Special Member
 
JohnAd's Avatar
 
Join Date: Jun 2000
Location: London
Posts: 1,725
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by dlarsen View Post

1) As noted here, more brightness and dynamic range over the valid signal range.

Broadly agree except for use of valid

Quote:
Originally Posted by dlarsen View Post

2) If one forsakes the notion of maintaining the BTB/WTW region, then the opportunity of using the full 8 bits over the valid range is open rather than using only 7.7 bits over the valid range.

I strongly disagree with this assertion, information cannot be created, an extra fractional bit of information does not appear magically by expanding the range. In fact noise is added by performing this transformation.

Note that this has nothing to do with were to set contrast on the display. Most of the valuable discussion here assumes that the display will be sent studio RGB and that the display will be set so that BTB is not visible. The outstanding question is where it is sensible to set the point at which WTW clips for digital displays.

Quote:
Originally Posted by dlarsen View Post

I know, the argument against expansion is banding. The argument goes that if you don’t maintain the monotonic nature of 1-254 then remapping and banding will occur.

No the monotonic nature is preserved, it is the smoothness of the gaps between the steps that it not preserved, most gaps stay at 1 but some jump from 1 unit to 2 units.

Quote:
Originally Posted by dlarsen View Post

3) If one forsakes the notion that valid BTB/WTW image information is encoded in all consumer material, and expands such that all 8 bits are use for the valid signal range, then total harmony with sRGB is achieved.

Agreed harmony would be nice. Personally I find going the other way, i.e using a gamma ramp to set all 0-255 material to be 16-235 to be preferable, assuming you have a 10bit output like on most ATI cards.


Quote:
Originally Posted by dlarsen View Post

Where is the banding?

Perhaps banding is a misleading term see above. I do see evidence of the smoothness issue. I see increased evidence of MPEG artifacts and noise in these images compared to what I would expect. A very good example of that would be the sky in the 5th picture. This looks extra noisy to me, this is consistent with the idea that pixels that were 1 unit apart in 16-235 space are now 2 apart and so more visible, this is to me the most obvious issue with the images and to some extent it is present on all of them. Whatever you want to call this effect I'd say it is visible. There may well be highlights clipped but I think that is a very secondary issue to the smoothness one.

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
post #194 of 484 Old 04-18-2010, 05:22 AM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 40
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by dlarsen View Post

I don't have a list, but most possible YCbCr triads map to an out-out-gamut or nonsensical color and are invalid. There are only ~2.8 million valid, in-gamut YCbCr triads. See Poynton.

That's true only when correlating 8bit YCbCr and 8bit RGB, but it's not a big problem, because all sources only have valid YCbCr triads (they are created from RGB)

Quote:
Originally Posted by dlarsen View Post

I know, the argument against expansion is banding. The argument goes that if you don't maintain the monotonic nature of 1-254 then remapping and banding will occur.

You can avoid the "banding" by dithering.

I think it's not very clear in this discussion to which clipping you are referring to. The 16-235 reference values are for YCbCr, but it seems some of you are looking for RGB values outside 16-235, and we should be looking for YCbCr values outside 16-235...
If we want to clip, we should clip in YCbCr, not in RGB. Personally, I always use YCbCr 16-245 -> RGB 0-255. With YCbCr 16-235 -> RGB 0-255 sometimes I get some crushed whites.
I will look for some images to help the discussion...
yesgrey is offline  
post #195 of 484 Old 04-18-2010, 06:28 AM
AVS Special Member
 
JohnAd's Avatar
 
Join Date: Jun 2000
Location: London
Posts: 1,725
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Here are some results from American Gangster. I've written a quick and dirty program to analyse the YUV output of each frame and look for number of pixels over peak and the maginitude of the peak.

It's only run over the first 3 minutes so far ... There did appear to be something odd in the universal logo so I've excluded that.

First table is number frames that have a given peak Y value

<=235 1657
236 7
237 12
238 11
239 6
240 15
241 18
242 18
243 17
244 13
245 9
246 6
247 10
248 14
249 10
250 9
251 4
252 1



So around 10% of frames have peak white over 235 with the highest at 252 at frame 3005

Largest number of pixels found over 235 so far is 2040 at frame 3008. These 2 peaks may have something to do with the use of burnt in titles so may not be that relevant, I'll wait until the film has processed before drawing too many conclusions.

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
post #196 of 484 Old 04-18-2010, 07:12 AM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by yesgrey View Post

The 16-235 reference values are for YCbCr, but it seems some of you are looking for RGB values outside 16-235, and we should be looking for YCbCr values outside 16-235...
If we want to clip, we should clip in YCbCr, not in RGB. Personally, I always use YCbCr 16-245 -> RGB 0-255. With YCbCr 16-235 -> RGB 0-255 sometimes I get some crushed whites.
I will look for some images to help the discussion...

Not sure I follow you. There should be zero clipping going on in the chain up till the display. At this point we are really only concerned with where to clip white in RGB referenced to the end display itself.

I suspect the only conclusion we can come to on this is to see what sort of level is actually hitting the screen. Its going to be 3 options.

254 mapped to peak white.
235 mapped to peak white.

some other number...

I was firmly in camp 1 and reassert that broadcast video displays disclose level all the through to 254.

However being a rational human being (okay maybe not that rational) I'm prepared to see where the numbers lie in the real world.

The aims are either :

We need to keep it all to 254 (on display) otherwise it will look clipped.

We can clip further down towards 235 to maximise contrast performance.

Either one will benefit presentation depending on what we find out.

digital film janitor
Mr.D is offline  
post #197 of 484 Old 04-18-2010, 07:17 AM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by JohnAd View Post



So around 10% of frames have peak white over 235 with the highest at 252 at frame 3005

Largest number of pixels found over 235 so far is 2040 at frame 3008. These 2 peaks may have something to do with the use of burnt in titles so may not be that relevant, I'll wait until the film has processed before drawing too many conclusions.

John

Great thanks John.

I'll check back in once my kids finish with Mama Mia

digital film janitor
Mr.D is offline  
post #198 of 484 Old 04-18-2010, 09:24 AM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 40
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by JohnAd View Post

So around 10% of frames have peak white over 235 with the highest at 252 at frame 3005

Thanks for the numbers. I will follow with some images...

Quote:
Originally Posted by Mr.D View Post

Not sure I follow you. There should be zero clipping going on in the chain up till the display. At this point we are really only concerned with where to clip white in RGB referenced to the end display itself.

Yes, but if we find that our sources exceed 16-235 only in RGB but not in YCbCr, we can improve the constrast by changing the YCbCr->RGB conversion.

Using mpc-hc, avisynth, and 3 custom 3DLUT files I've created, here are some images to show how much <16 and >235 data exists in a movie.
I've used it with the first 1m20s of the Blu-ray Babylon.
The zip file I linked contain 2 frames of the movie. One with a near reference black image (Babylon007), and another with a near reference white image (Babylon011).
The images without any extension are without any masking.
The images with "RGB_16-235_GR" in the name show pixels with RGB < [16,16,16] at green, and pixels with R>235 or G>235 or B>235 at red.
The images with "Y_16-235_GR" in the name show pixels with Y < 16 at green, pixels with Y>235 at red, and pixels with CbCr outside the [16-240] range at yellow.

As we can see in image 7, there are a lot of pixels < 16 in both RGB and YCbCr, but when we look at the higher level side (using image 11), > 235, eventhough there are a lot of pixels with R,G or B > 235, there aren't any with Y >235...

The images are PNGs at native resolution, so I zipped them and uploaded the pack here.
yesgrey is offline  
post #199 of 484 Old 04-18-2010, 10:29 AM
AVS Special Member
 
JohnAd's Avatar
 
Join Date: Jun 2000
Location: London
Posts: 1,725
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by yesgrey View Post

Thanks for the numbers. I will follow with some images...

It still running, it's at 11mins in now....

Here is a possibly more interesting section, with no titles

frame countYOver235 Ypeak
6301 34 244
6302 50 245
6303 35 241
6304 95 246
6305 234 246
6306 242 244
6307 243 252
6308 259 245
6309 192 244
6310 214 246
6311 242 243
6312 275 246
6313 360 244
6314 337 245
6315 127 245
6316 119 243
6317 204 247
6318 236 247
6319 238 244
6320 457 244
6321 562 245
6322 541 243
6323 272 244
6324 119 245
6325 32 245
6326 53 242
6327 23 240
6328 14 240
6329 6 237
6330 5 237
6331 3 238

(Note frame is zero based)

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
post #200 of 484 Old 04-18-2010, 10:34 AM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by yesgrey View Post


As we can see in image 7, there are a lot of pixels < 16 in both RGB and YCbCr, but when we look at the higher level side (using image 11), > 235, eventhough there are a lot of pixels with R,G or B > 235, there aren't any with Y >235...
.

I get that but we are primarily interested in where white should be on display. That means RGB ....notionally Y doesn't come into it.

I really think we should avoid getting into areas other than categorising where people should set the contrast control on their displays. Forget YCrCb lets just see where we should clip RGB at the display and work back from there.

I'm still sat here watching "Mama Mia" and now "Nacho Libre" (kids are too young for District 9) with my display clipping out at 254 and I'm having a hard time switching over to clipping below that on the display even with the suspicion that there is little above 240,

digital film janitor
Mr.D is offline  
post #201 of 484 Old 04-18-2010, 10:46 AM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by JohnAd View Post

It still running, it's at 11mins in now....

Here is a possibly more interesting section, with no titles

frame countYOver235 Ypeak
6301 34 244
6302 50 245
6303 35 241
6304 95 246
6305 234 246
6306 242 244
6307 243 252
6308 259 245
6309 192 244
6310 214 246
6311 242 243
6312 275 246
6313 360 244
6314 337 245
6315 127 245
6316 119 243
6317 204 247
6318 236 247
6319 238 244
6320 457 244
6321 562 245
6322 541 243
6323 272 244
6324 119 245
6325 32 245
6326 53 242
6327 23 240
6328 14 240
6329 6 237
6330 5 237
6331 3 238

(Note frame is zero based)

John


So we are definitely still talking level occasionally above 250.

digital film janitor
Mr.D is offline  
post #202 of 484 Old 04-18-2010, 10:56 AM
AVS Special Member
 
JohnAd's Avatar
 
Join Date: Jun 2000
Location: London
Posts: 1,725
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by Mr.D View Post

So we are definitely still talking level occasionally above 250.

yes seems so, but the quantity of information there is very low. I'll run my thing over a few more films, see what I get.

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
post #203 of 484 Old 04-18-2010, 12:26 PM
AVS Special Member
 
JohnAd's Avatar
 
Join Date: Jun 2000
Location: London
Posts: 1,725
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by JohnAd View Post

It still running, it's at 11mins in now...

Given that it was running so slowly I've restarted it with release mode code and some extra code to dump out tiffs of any interesting frames (either count or peak over 250).

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
post #204 of 484 Old 04-18-2010, 01:01 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 40
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by JohnAd View Post

Here is a possibly more interesting section, with no titles

Hmmm, this shows that I pointed almost to the right spot with my YCbCr at 16-245...

Quote:
Originally Posted by Mr.D View Post

I get that but we are primarily interested in where white should be on display. That means RGB ....notionally Y doesn't come into it.

Yes, but the white depends on the conversion from YCbCr to RGB. I only agree not to consider it because only people with a HTPC could change it, so let's "forget it", for now, just to simplify this.

Quote:
Originally Posted by Mr.D View Post

I really think we should avoid getting into areas other than categorising where people should set the contrast control on their displays. Forget YCrCb lets just see where we should clip RGB at the display and work back from there.

But for people with a HTPC it would be preferable the other way around...

Quote:
Originally Posted by Mr.D View Post

with my display clipping out at 254 and I'm having a hard time switching over to clipping below that on the display even with the suspicion that there is little above 240,

People are forgetting another thing about the clipping: the display's gamma curve.
Some display's have a hard time at the extremes, hence why we need some headroom at the top, even if there are no values in the source.
My projector's gamma curve near the top shows some kind of "saturation" effect. So, If I clip at lower values, without any headroom, the whites will look crushed. Not because I am missing video data that was on the source, but because the white levels are less far apart than they should...
Maybe that's one of the reasons for you to dislike clipping at a lower value...
yesgrey is offline  
post #205 of 484 Old 04-18-2010, 01:46 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by yesgrey View Post

Hmmm, this shows that I pointed almost to the right spot with my YCbCr at 16-245...


Yes, but the white depends on the conversion from YCbCr to RGB. I only agree not to consider it because only people with a HTPC could change it, so let's "forget it", for now, just to simplify this.


But for people with a HTPC it would be preferable the other way around...


People are forgetting another thing about the clipping: the display's gamma curve.

Lets assume even if you use an HTPC you want it passing "unmolested" values before you introduce a CMS.

To my mind gamma should by definition be a totally seperate issue to white point. Like I said all we are currently concerned about is where peak white in the material should be mapped to the display. All other considerations are by the by .

digital film janitor
Mr.D is offline  
post #206 of 484 Old 04-18-2010, 02:32 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 40
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

To my mind gamma should by definition be a totally seperate issue to white point.

I totally agree.
I only referred the gamma in an attempt to justify your disliking of the clipping at a lower value...
yesgrey is offline  
post #207 of 484 Old 04-18-2010, 03:14 PM
AVS Special Member
 
Mr.D's Avatar
 
Join Date: Dec 2001
Location: UK
Posts: 3,307
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 15
Quote:
Originally Posted by yesgrey View Post

I totally agree.
I only referred the gamma in an attempt to justify your disliking of the clipping at a lower value...

Fair enough I'm just concerned that we don't get off track. Its one of those issues that needs clarification for everyone's sake.

digital film janitor
Mr.D is offline  
post #208 of 484 Old 04-18-2010, 03:34 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by Mr. D View Post

Lets assume even if you use an HTPC you want it passing "unmolested" values before you introduce a CMS.

I've never really wrapped my head around this either. It's preferable to molest' on one end of the cable, but not the other. If I'm going to molest, I'd prefer to do it where I have the most precision and control.

As John noted earlier, I can't create information' from nothing, but as yesgrey noted, I can effectively dither' a bit with a smaller spoon when I spoon it out from the normalized soup bowl where is begins the transformation. The first step in the YCbCr transform calls for a normalization of the valid YCbCr range to [0..1]. I can get smaller bites (smaller steps between values) by dipping into the normalized soup with a 1/256 spoon than I can with a 1/220 spoon. Yes, if I don't dip into it with the pure 1/220 size spoon, I might get a 1.7X lump, but this can only happen when the CbCr values aren't weighted in the transform. (About 1540 times out of the 2.8 million normalized [0..1] YcbCr values) For all the other real-world (non test-pattern cases where CbCr does weight the transform) image cases, my spoon and step size can be 15% smaller so should have less chance of rising above my ability to detect a step. Seems to me anyway.

If I do the molestation' to the [0..1] normalized YcbCr soup by changing the gains and offsets applied there, I should be doing it at 32/64 bits and/or with floating point precision. Changing or molesting it with 32/64 or floating point precision seems to me to be the best place to molest it. I'm gonna molest it somewhere or I wouldn't calibrate it. If I change gains/offsets only in my display, I may be doing it with less than 8 or 10 bit precision. I think my CPU is capable of much more precision than my procamp controls are.

Apologies for going off reservation, but this also address some comments that were directed specifically at me earlier here. I probably will respond if I'm prompted to.

Quote:
Originally Posted by yesgrey View Post

Hmmm, this shows that I pointed almost to the right spot with my YCbCr at 16-245...

245. Split the difference.

Dave
dlarsen is offline  
post #209 of 484 Old 04-18-2010, 05:21 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 40
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

Its one of those issues that needs clarification for everyone's sake.

Yes, it's time for someone to do it.
Would it be useful if JohnAd also performed the countig for black level (below 16)? Or do we all agree with the idea of clipping below 16 (I do)?

Quote:
Originally Posted by dlarsen View Post

245. Split the difference.

It's always a good start...
yesgrey is offline  
post #210 of 484 Old 04-18-2010, 05:58 PM
AVS Addicted Member
 
millerwill's Avatar
 
Join Date: Apr 2004
Location: Berkeley, CA
Posts: 11,381
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 21 Post(s)
Liked: 29
Quote:
Originally Posted by yesgrey View Post

Would it be useful if JohnAd also performed the countig for black level (below 16)? Or do we all agree with the idea of clipping below 16 (I do)?

Yes, I think everyone agrees on BTB; so PLEASE, let's don't open up this can of worms.
millerwill is offline  
Reply Display Calibration

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off