Setting contrast with Spears and Munsil disc - Page 16 - AVS Forum
Forum Jump: 
 
Thread Tools
Old 05-08-2010, 05:37 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by madshi View Post

Color space conversion matrices are sometimes written with signed int CbCr values in mind, though. And I find that more intuitive, anyway. Having 0 as neutral (colorless) just makes much more sense than having 0x80 as neutral.

Just a little extra info.
CbCr are always unsigned. The signed CbCr you are referring is not CbCr, is PbPr.
CbCr is derived from PbPr by adding an offset value to make all values >=0.
yesgrey is offline  
Sponsored Links
Advertisement
 
Old 05-08-2010, 06:00 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

This is the crux of the matter. When I grade I don't explicitly aim for 235 as the white point. My rec.709 environment shows detail all the way to 254. Most colorists I've seen if grading subjectively will not be trying to hit a number, they will grade the material to look subjectively pleasing within the definitions of their reference environment.

When you say "When I grade I don't explicitly aim for 235 as the white point" are you referring to RGB 235 or YCbCr 235?
One thing is during the grading in RGB the white point being considered as 255 and the black point as 0, but then, when converting to YCbCr, consider the YCbCr white point as 235 and the black point as 16. If it's done this way, then there will be very few data above 235 and below 16 in YCbCr but in RGB there will be significant data above those values "if" the correct inverse matrix is used for the conversion back to RGB.

Are there anyone grading using 16-235 RGB?

And how did you create your test images? 16-235 RGB and 16-235 YCbCr or 0-255 RGB and 16-235 YCbCr?
yesgrey is offline  
Old 05-10-2010, 02:46 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 yesgrey View Post

When you say "When I grade I don't explicitly aim for 235 as the white point" are you referring to RGB 235 or YCbCr 235?

RGB

Quote:
Originally Posted by yesgrey View Post

One thing is during the grading in RGB the white point being considered as 255 and the black point as 0, but then, when converting to YCbCr, consider the YCbCr white point as 235 and the black point as 16. If it's done this way, then there will be very few data above 235 and below 16 in YCbCr but in RGB there will be significant data above those values "if" the correct inverse matrix is used for the conversion back to RGB.

Yes but what we are investigating is what exactly the picture detail is above 235. Certainly from looking at BD the vast majority of the data seems to be noise generated from resizing and resampling: or at least it could be that we seem to be able to replicate ie: imagery with no level above 235 prior to resizing and resampling has pixels created that reach all the way to 255: this replicates what we see from BD.

That coupled with the fact that when we clip the BD images to 235 the visual impact seems so minute as to make it pretty much impossible to differentiate what was a real pixel from generated noise.

Generally we see level above 235 in RGB in most cases but the actual pixels above 235 look like noise clustered around abrupt luma and chroma transitions (edges in other words). It rarely if ever seems to behave as real photographic picture detail ( see Almost Famous image and look at the flare on the car bonnet).


Quote:
Originally Posted by yesgrey View Post


Are there anyone grading using 16-235 RGB?

I don't know if anyone is specifically aiming for a 235 maximal whitepoint or its there as a result of the pipeline or some legalising process after the fact. They won't notionally be grading in 8bit at all , probably 16bit integer at least , more likely 32bit float.

The key thing is are they notionally aware of where their level is ...and what it actually is. I suspect they mostly just rely on the display environment and their subjective "eye" to not place white at the extremes of the range.

For a modern film its quite likely that the bulk of the transfer from film to video is mostly handled with a one size fits 99% transform to mimic the print look into a rec.709/video environment. In which case its a question of how the transform is designed.

This is actually what I think is happening : film graded as 10bitlog data in a print display environment with 32 bit float precision. Graded data converted to video with transform to replicate film look in video target. Scenes that are compromised too much then retweaked by eye ( probably more like they have a handful of luts that they throw if the image is too dark/too clipped etc as the film grade is doing the bulk of the work in conjunction with the default transform).

For legacy films they may not bother with a film grade although the baseline that they initially work to may be totally immaterial in terms of prior to hitting video.

However they actually do it figuring out where they are actually putting white is the main thing. Certainly the BD data seems to suggest that 235 is the actual white limit with subsequent level above this point highly suggestive of consisting of processing noise.

Quote:
Originally Posted by yesgrey View Post


And how did you create your test images? 16-235 RGB and 16-235 YCbCr or 0-255 RGB and 16-235 YCbCr?

Image created in 16bit with 16-235 RGB limits set with reference to 8bit code value (colorwheel actually hits 236 probably a rounding error from the 16bit origination) mapped into YCrCb, luma resized to 1080p , chroma resized to 480x270 to replicate 4:2:0 subsampling with a variety of filters , conversion back to RGB , bitdepth dropped to 8bit from 16bit , written out as PNG.

I have located a 2k fullap marcie. So I'll do the chroma subsampling and resize on a real photographic image just as soon as I can breathe some life back into the ancient W2K workstation its sat on.

That will be 10bit log RGB , visually graded into video with a 235 hard limit , converted to YCbCr , resize to 1080p and 4:2:0 downsample (also do a 1080p extraction rather than resize , thats why I wanted a fullap scan) upsample and conversion back to RGB for viewing. See if the range over 235 looks similar to the BD material.

I do have some 10bit log material from feature films ( actually I have more of this than video) that I can compare directly to the BD version. Its how I came to the conclusion that 1080p masters seem to be quite often 1080p extractions from the 2k without any resizing. However I will not post or distribute it so you would kinda have to take my word for any findings...which is problematic.

digital film janitor
Mr.D is offline  
Old 05-10-2010, 03:10 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

As a quick round up here is where I think we are for hi-def discs

Mr.D has run (at least):
Transformers 2
Avatar
Pandorum


Results:

All content tested has some pixels over 235 in both YUV and RGB

Add "Fantastic Mr. Fox" and "The New World".

Not seeing anything that contradicts current findings. In fact Mr.Fox is interesting as the bulk of the frames with level over 235 are the end credits with orange text against black (sharp edges across a large area of image) or the 20th Century Fox logo at the start. (sharp geometric artificial imagery).

Again images that you would expect to have level above 235 (lights directly into camera all seem to top out at 235 in thos areas , its only the bright edge transitions that hit above 235)

digital film janitor
Mr.D is offline  
Old 05-10-2010, 04:00 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

Add "Fantastic Mr. Fox" and "The New World".

Thanks, done.

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
Old 05-10-2010, 12:01 PM
AVS Special Member
 
bodosom's Avatar
 
Join Date: Aug 2002
Location: Niagara Frontier
Posts: 6,395
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 22 Post(s)
Liked: 22
[This is bit off-track but peripherally related]

In his recent column about gamma Poynton asserts that final approval is now done in the near dark contrary to the old SMPTE RP guidance. 100 nt/1 lx/1% versus 100 nt/6 lx/10%. I wonder if any colorists watching care to offer any commentary.

Certainly this would bear on setting "contrast".
bodosom is offline  
Old 05-14-2010, 02:21 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 bodosom View Post

[This is bit off-track but peripherally related]

In his recent column about gamma Poynton asserts that final approval is now done in the near dark contrary to the old SMPTE RP guidance. 100 nt/1 lx/1% versus 100 nt/6 lx/10%. I wonder if any colorists watching care to offer any commentary.

Certainly this would bear on setting "contrast".

Generally speaking all the film DIs I've been in have a dark environment with a projector. Not sure if they bother changing it for video appraisal. They may lock the white point for film work ( I normally do if I'm using a CRT) and let it scale for video. The last video grading suite I was in had a fairly normal dim environment (not pitch black) but I wasn't terribly impressed with the display setup itself anyway. it wasn't accurate enough for me.

I'll be doing the additional 2k film (actually 4k) to 1080p 4:2:0 tests this weekend : again it will be a case of seeing what sort of pixels are created over 235 as a consequence of the resizing and chroma sampling stages. And whether this tallies with what we see from the BD examples.

I will also see if I can verify if anything above 235 exists on a shot from BD that actually contains any real picture detail relative to the original film scan.

(frame grab thats confirmed as setup for capturing level from BD all the way to 254 if its there and whether any pixels above 235 represent real detail or noise...I won't be able to post any examples from this...its a recent blockbuster though)

digital film janitor
Mr.D is offline  
Old 05-14-2010, 08:38 AM
AVS Addicted Member
 
millerwill's Avatar
 
Join Date: Apr 2004
Location: Berkeley, CA
Posts: 11,444
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 58 Post(s)
Liked: 47
Look forward to your report, Mr. D. And thanks again for all your efforts on this topic.
millerwill is offline  
Old 05-16-2010, 04:52 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
I'll add to this post throughout the day.

Input is a 4k (3656x2664 academy) 10bit log Kodak LAD Chart ("marcie" even though thats not her name): many thanks to my friends at Kodak for providing it.

pipeline is: 4k 10bitlog 32bit float workflow , color correction to video (eyeballed by me) full range of 1-1023 mapped into 1-235. Nothing above 235 in 8bit code values , padded to 4096x3112 (fullap) cropped to 4096x2204 (this is to keep the 1080p resize integer for our purposes). linearisation (just by applying a simple 2.2 gamma) resize to 1080p with triangle filter and inverse the linearisation : output then cropped to meaning area and deflate to 8bit.

This is arguably softer looking than I'd say is acceptable however it is as clean as I can make it in terms of ringing and has no resulting pixels generated over 235 by the resize (the uploading has turned the png into a jpg again which has added artififacts so its for illustration only)




4k log to 1080p 4:2:0 video with triangle filters for all scaling stages. No level above 235 generated.



4k log to 1080p 4:2:0 video with since filter on Y scale to 1080 chroma is still triangle: level generated above 235.



pixels above 235


LL
LL
LL
LL

digital film janitor
Mr.D is offline  
Old 05-16-2010, 05:01 AM
AVS Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 5,482
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 70 Post(s)
Liked: 140
Clean of ringing? I see very strong ringing in the colored squares! Maybe the forum software has done violence to your uploaded image? It might be a better idea to use one of those free image hosts.

http://www.findimagehost.com/unlimit...ge-hosting.php
madshi is offline  
Old 05-16-2010, 06:18 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 madshi View Post

Clean of ringing? I see very strong ringing in the colored squares! Maybe the forum software has done violence to your uploaded image? It might be a better idea to use one of those free image hosts.

http://www.findimagehost.com/unlimit...ge-hosting.php

All in good time madshi I also have a roast chicken on the go.

Its clean (thats why I didn't bother to isolate the value above 235 as there isn't any) the forums jpegging has caused the artifacts. It also has no chroma subsampling.

digital film janitor
Mr.D is offline  
Old 05-16-2010, 12:25 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
4k log to 1080p 4:2:0 video with sinc filter on Y scale to 1080 chroma is sinc on the downsample , upsample still triangle: level generated above 235.



pixels above 235



sinc on both Y and chroma scaling stages: level above 235



pixels above 235




The fringing on the blocks is not quite as bad as the uploads appear here but there are quite visible artefacts , I'll see about getting the pngs uploaded and swapped in when I get some time.
LL
LL
LL
LL

digital film janitor
Mr.D is offline  
Old 05-16-2010, 12:50 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
Now something a little bit different. The previous examples were based around 4k to 1080p workflow. Most of the time we will be getting 4k downsampled to 2k log with a sinc filter of some description and then a further downsize of the 2k to to 1080p. I suspect that its since all the way apart from the chroma subsampling.

4k to 2k (sinc) to 1080p sinc ,triangle on chroma upsample. all rescales in linear.



pixels above 235



This is the same without bothering to go linear for the scaling stages:



level over 235


LL
LL
LL
LL

digital film janitor
Mr.D is offline  
Old 05-16-2010, 01:05 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
Conclusions,

You can generate level above 235 with rigorously mastered material from the flitering used in the scaling and chroma subsampling stages. We knew this from the previous round of pattern tests anyway.

It looks to me to be fairly similar to the greater than 235 level we are seeing on BD. The noise generated in the hair could well be mistaken for WTW detail at least on a single frame.

The less aggressive filters (triangle) don't have this effect however they look visually soft. I for one would probably still prefer the sinc examples albeit with a triangle for the chroma downsample. The blocks on the LHS although they look horrendous are worst case scenario and should be appraised as such nothing more nothing less.

I do suspect that the video versioning does not happen in linear most of the time, nor does the chroma subsampling and its more often than not a case of 4k to 2k to 1080p with the possibility of a crop to 1080p from the 2k rather than another resize.

So with all that in mind what does it tell us about contrast?

Here's my opinion:

Video is generally mastered aiming for a 235 whitepoint, in the cases where its not rigorously enforced the level rarely gets up much above 240. Think of it as an accuracy tolerance (wiggle room). In most cases the vast majority of pixels above 235 are in fact noise generated along the way and not real picture info.

We should set contrast a little bit above 235 say to 240 to allow for the possibility of minor inaccuracy at mastering.

Now I would like to hear some rebuttal regarding any possible reasons as to why its bad to clip lower than 254 on display other than to preserve real picture information because I'm pretty satisfied there is nothing much up there.

digital film janitor
Mr.D is offline  
Old 05-16-2010, 01:36 PM
AVS Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 5,482
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 70 Post(s)
Liked: 140
Quote:
Originally Posted by Mr.D View Post

Here's my opinion:

Video is generally mastered aiming for a 235 whitepoint, in the cases where its not rigorously enforced the level rarely gets up much above 240. Think of it as an accuracy tolerance (wiggle room). In most cases the vast majority of pixels above 235 are in fact noise generated along the way and not real picture info.

We should set contrast a little bit above 235 say to 240 to allow for the possibility of minor inaccuracy at mastering.

Now I would like to hear some rebuttal regarding any possible reasons as to why its bad to clip lower than 254 on display other than to preserve real picture information because I'm pretty satisfied there is nothing much up there.

Are we talking about DVD / Blu-Ray, only? I thought there was more WTW in broadcasts? Or did I get that wrong?
madshi is offline  
Old 05-16-2010, 02:14 PM
Member
 
cyberlolo's Avatar
 
Join Date: Jan 2010
Posts: 126
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

Conclusions,

Here's my opinion:

Video is generally mastered aiming for a 235 whitepoint, in the cases where its not rigorously enforced the level rarely gets up much above 240. Think of it as an accuracy tolerance (wiggle room). In most cases the vast majority of pixels above 235 are in fact noise generated along the way and not real picture info.

We should set contrast a little bit above 235 say to 240 to allow for the possibility of minor inaccuracy at mastering.

Now I would like to hear some rebuttal regarding any possible reasons as to why its bad to clip lower than 254 on display other than to preserve real picture information because I'm pretty satisfied there is nothing much up there.

Do this recommendations apply to YCbCr or to RGB output?
cyberlolo is offline  
Old 05-16-2010, 05:15 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 cyberlolo View Post

Do this recommendations apply to YCbCr or to RGB output?

This is all about where the detail ends so we are talking RGB. Where white appears visually.

Its not really an issue of what sort of output you use, it all ends up as RGB at your eyeball.

digital film janitor
Mr.D is offline  
Old 05-16-2010, 05:29 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 madshi View Post

Are we talking about DVD / Blu-Ray, only? I thought there was more WTW in broadcasts? Or did I get that wrong?

What we seem to be seeing on broadcast is that in certain situations that are not as notionally controlable such as live reports/sports etc., there may be more detail above 235. However more carefully mastered stuff seems to top out at 235.

In most cases the detail that does go above 235 is in areas that are so blown out that they look clipped regardless of whether we keep it all to 254 or just lop some off to improve the contrast. This is one area that we haven't spent as much time on with hard analysis.

I am running TV broadcasts here with white at 235, 240 and 254 and not seeing any improvement to the blown out areas at 254.
In most instances where I think its a bit clippy looking dropping back to 254 doesn't visually improve things but does incurr a perceptable loss of contrast.

The 240 point seems to be a good place in terms of maximising contrast and providing some leeway for small excursions. I'm pretty happy with BD at 235 , the 240 suggestion is just to provide a little margin for error.


As ever try it and see what you think.

digital film janitor
Mr.D is offline  
Old 05-16-2010, 06:26 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

pipeline is: 4k 10bitlog 32bit float workflow , color correction to video (eyeballed by me) full range of 1-1023 mapped into 1-235. Nothing above 235 in 8bit code values

Is it possible for you to repeat this but pointing to RGB 1-254 instead of 1-235?
It would be interesting to know if the lack of overshoot zone will affect the resulting image quality after all the resizing and chroma subsampling...

Quote:
Originally Posted by Mr.D View Post

Now I would like to hear some rebuttal regarding any possible reasons as to why its bad to clip lower than 254 on display other than to preserve real picture information because I'm pretty satisfied there is nothing much up there.

The only reason I can come up is if the display has a really bad gamma curve at its top. Not clipping lower than 254 would use only the good portion of the gamma curve, because no data would use the bad part. Though, I think 254 might be too much...
yesgrey is offline  
Old 05-17-2010, 02:34 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

Is it possible for you to repeat this but pointing to RGB 1-254 instead of 1-235?
It would be interesting to know if the lack of overshoot zone will affect the resulting image quality after all the resizing and chroma subsampling...

Can do. Although it will be tricky to isolate any resulting noisy pixels from the scaling. I can probably do a triangle version and subtract it but that will only show the difference between sinc and triangle which we kinda know about already.

Quote:
Originally Posted by yesgrey View Post

The only reason I can come up is if the display has a really bad gamma curve at its top. Not clipping lower than 254 would use only the good portion of the gamma curve, because no data would use the bad part. Though, I think 254 might be too much...

I'm making the assumption that gamma is notionaly perfect on the display (it is pretty much "perfect" on mine courtesy of John Adcock's Upsilon Mixer CMS I'm using on my HTPC). I can see your point though depending on display capabilities.

digital film janitor
Mr.D is offline  
Old 05-17-2010, 09:22 AM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

Can do. Although it will be tricky to isolate any resulting noisy pixels from the scaling.

But I don't want you to isolate the noisy pixels, I want you to post both images to see if we can notice any problems in the final image... but for that we would need the images in png instead of jpg...
yesgrey is offline  
Old 05-17-2010, 10:02 AM
AVS Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 5,482
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 70 Post(s)
Liked: 140
Quote:
Originally Posted by Mr.D View Post

I'm making the assumption that gamma is notionaly perfect on the display (it is pretty much "perfect" on mine courtesy of John Adcock's Upsilon Mixer CMS I'm using on my HTPC).

Just read up on JohnAd's Mixer CMS software. Pretty interesting. yesgrey and I are working on something similar (slightly different concept, though). Seems that JohnAd and we will be competitors...
madshi is offline  
Old 05-18-2010, 02:51 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 madshi View Post

Just read up on JohnAd's Mixer CMS software. Pretty interesting. yesgrey and I are working on something similar (slightly different concept, though). Seems that JohnAd and we will be competitors...

Thats all good in my opinion , the more activity in this area the better. I really feel that HTPCs have the potential to be the pinnacle of video sources just the implimentation that has so far been lacking. I do feel that the HTPC scene has lost its way in recent years.

Too many companies putting together just barely functional video apps usually with developers who don't know or care about good video quality. Then you throw badly thought out drivers into the mix. Considering the amount of video processing built into graphics cards these days you would think that the hardware developers would make sure it could be harnessed properly. Especially as they made a big deal out of it a few years back.

I did like your madVR but could never get it to run without stuttering, might have been something else will give it another bash at some point.

Grading for professional film and video production happens on fairly unremarkable workstations these days. No reason why an HTPC can't be made to work at the same level with full CMS and correctly implimented processing. I'm very happy to see John back in the saddle on this.

digital film janitor
Mr.D is offline  
Old 05-18-2010, 03:09 AM
AVS Special Member
 
madshi's Avatar
 
Join Date: May 2005
Posts: 5,482
Mentioned: 9 Post(s)
Tagged: 0 Thread(s)
Quoted: 70 Post(s)
Liked: 140
Quote:
Originally Posted by Mr.D View Post

Thats all good in my opinion , the more activity in this area the better. I really feel that HTPCs have the potential to be the pinnacle of video sources just the implimentation that has so far been lacking. I do feel that the HTPC scene has lost its way in recent years.

Too many companies putting together just barely functional video apps usually with developers who don't know or care about good video quality. Then you throw badly thought out drivers into the mix. Considering the amount of video processing built into graphics cards these days you would think that the hardware developers would make sure it could be harnessed properly. Especially as they made a big deal out of it a few years back.

Agreed. Very much so.

Quote:
Originally Posted by Mr.D View Post

I did like your madVR but could never get it to run without stuttering, might have been something else will give it another bash at some point.

madVR has also been quite unstable (crashes and freezes) for a while. This has finally been fixed in the latest version, though. Stuttering is usually due to a too slow GPU. madVR needs a GPU with a reasonably high pixel shader power.
madshi is offline  
Old 05-18-2010, 03:36 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 madshi View Post

Seems that JohnAd and we will be competitors...

friendly ones I hope.

As Mr.D puts it above it feels like time to give those GPU's a proper work-out.

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
Old 05-18-2010, 03:54 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

Now I would like to hear some rebuttal regarding any possible reasons as to why its bad to clip lower than 254 on display other than to preserve real picture information because I'm pretty satisfied there is nothing much up there.

The argument for keeping it would be to do with the fact that clipping alters the nature of the noise added by the resampling/chroma process. If you keep all the way to 254 then the total "weight" of the image is as intended, as the noise is both above and below, by chopping off the only the above part of the noise you get left with potentially more objectionable asymmetric noise.

Personally I think the practical visible effect of clipping is so small that the weighting given to keeping the full range against other considerations should be small.

John

My Company - Upsilon Software
Free Projects - DScaler & hcfr fork
JohnAd is offline  
Old 05-18-2010, 06:05 AM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

Now I would like to hear some rebuttal regarding any possible reasons as to why its bad to clip lower than 254 on display other than to preserve real picture information because I'm pretty satisfied there is nothing much up there.

Another reason to not clip would be for preserving the original transfer function (gamma) of the video. If we clip at the display level, the transfer function would not be exactly what it should be, so we would not see an accurate image. Though, we can always correct the display transfer function to compensate for the effect due to clipping...
yesgrey is offline  
Old 05-18-2010, 06:11 AM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by JohnAd View Post

friendly ones I hope.

yesgrey is offline  
Old 05-19-2010, 02:53 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

Another reason to not clip would be for preserving the original transfer function (gamma) of the video. If we clip at the display level, the transfer function would not be exactly what it should be, so we would not see an accurate image. Though, we can always correct the display transfer function to compensate for the effect due to clipping...

I think in that case it would be good to quantify what actually happens to the gamma when above 235 is clipped.

I'm assuming the end gamma result increases. (simply put) This in itself might not be a bad thing for many people as generally display gamma tends to be too low rather than too high.

digital film janitor
Mr.D is offline  
Old 05-19-2010, 05:01 PM
Member
 
yesgrey's Avatar
 
Join Date: Apr 2010
Posts: 44
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
Quote:
Originally Posted by Mr.D View Post

I think in that case it would be good to quantify what actually happens to the gamma when above 235 is clipped.

So we need to measure the gamma curves in both situations and compare...

Quote:
Originally Posted by Mr.D View Post

I'm assuming the end gamma result increases. (simply put) This in itself might not be a bad thing for many people as generally display gamma tends to be too low rather than too high.

Yes, but people with their display's gamma corrected might need a new calibration...
yesgrey is offline  
 
Thread Tools


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