Screencaps: What are they good for?

[Joe Cain]

(Posting in Off-Topic because the subject doesn't really seem to fit elsewhere; I trust the mods to relocate the thread if the subject or discussion seems useful enough to deserve wider play.)

Screencaps are being used a lot. I don't know much about the underlying technology. I've gotten the impression that the 1080 screencaps posted with the reviews here at blu-ray.com involve something proprietary---I'm not looking for that level of information.

But with last year's Dracula brouhaha, this year's redux with the Godfather trilogy, and other such discussions, there are some things I'd like to understand....things which have nothing to do with the "right"-ness of a given image, but which are more about what constitutes a quality capture, what information can fairly be deduced from such an image, and what criteria must be met for two captures to provide a fair basis for comparison and contrast.

If a full resolution capture is made of a BD or DVD's still image, is it necessarily accurate? Is the accuracy dependent on hardware or software? If a capture is down-resolved, whether as a part of the capture or afterwards as part of exporting a smaller file for posting online, what happens to the hue, the contrast, the sharpness? If, for the sake of comparing the same image from two separate releases, separate screen captures are made by different people on different hardware, does there exist a fair basis for comparison? Would that answer depend on whether or not the same capture process was employed by both parties? (And if both captures were less than full resolution, could a fair comparison be made of anything besides framing?)

Thanks in advance for useful, informed responses.

[dadkins]

"Computer" screen caps are data, not what is being displayed on the capture computer's display - like what would be a image taken with a camera.



Click for full size.

This might/could look green on this computer's display - I doubt it looks green on your screen.
No, it doesn't look green, just used as an example.

This is Kirby:



He's orange.
This is what was stored in memory when I took it.
I use JPEG at 100%.
I could use BITMAP but the files would be slightly too big for Imageshack.

Looks the same on this machine as it did on the other machine it was captured on.

With using a camera for "screen shots", you have more things that may affect what gets posted - the screen displaying the image, the camera taking the picture, and then whatever compression.

The above images are 1920x1080. Same they were played back as. No resizing was involved.
Not a camera image.
If you can get these to your HDTV, they would look great.
Other storage formats might be truer to actual images, but there is a point where the human eye will not be able to detect the difference between the actual displayed Blu-ray image and a static image on a memory chip.

[Anthony P]

what information can fairly be deduced from such an image

none. that is the joke and why people like PM make fun of them (not just the people that take screen caps but even more the people that use them to judge stuff.)

1) most people don't have their PC connected to their main display
2) most people don't have the PCs monitor/input properly calibrated or at the same settings for the TV/ BD player

so how can you judge something as simple as colour or brightness when both are 100% dependent on the display?

3) you need to understand codecs, codecs (all of them) work in time and space, there what are called I, B and P frames. an I frame represents the whole picture (each detail and all 1080x1920 pixels) because of this it is severely limited (i.e. it cannot contain all the detail in the pic) and so loses a lot of detail, for the next frame it will either be a b or p frame and it won't try to show the whole 1080x1920 it will just try and show the changes, these changes are either real changes (i.e. the leg part of the frame changed in the person walking but all the background is 100% the same) or what I call feeling in the blanks (i.e. it adds the crows feet in the close-up of the face which could not be represented in the previous i frame). So if someone captures the original i frame (which is there for 1/24th of a second), it will appear that there is heavy DNR but it is just that the detail has not been added yet with the next p or b frame. I frames happen when there is a large change (i.e. seeing what is happening somewhere else, image from a different camera angle....) and on a regular basis. One more thing is that since none i frames don't have all the detail and are relative with each consecutive none i frame there is a possibility of more issues (i.e. the decoder made a mistake when calculating what pixels are changed)

4) compression, for the most part people don't post uncompressed screen captures, the minute you use jpeg or gif or anything else you add artefacts from these codecs

[4K2K]

what information can fairly be deduced from such an image

Originally Posted by Anthony P

none. that is the joke and why people like PM make fun of them (not just the people that take screen caps but even more the people that use them to judge stuff.)

In my view it does give a rough idea of what it will be like, otherwise there would be no point in having them at all. It won't let you know what it looks like in motion, but it will give you a rough idea of what the picture quality is like and what is on the Blu-ray disc (in case you hadn't seen that title before).

So if someone captures the original i frame (which is there for 1/24th of a second), it will appear that there is heavy DNR but it is just that the detail has not been added yet with the next p or b frame.

I thought I frames were the most accurate frames - anyway I don't think it should really matter whether they capture an I, P or B frame as all frames should be encoded with high quality (high enough bitrate). The eye can see at over 60fps (or they wouldn't be releasing 120hz and 240hz LCD HDTVs), so I think problem frames at 24fps could still sometimes be visible. But you are correct, it won't let you know what it is like in motion, but it will give you an idea what the quality is like (and also possibly let you know if too much processing has been done to the picture).

[Anthony P]

In my view it does give a rough idea of what it will be like, otherwise there would be no point in having them at all.

d'uh, so you think because someone posts useless screen shots it means they can't be useless? what good is " a rough idea " when that rough idea can be 100% wrong (not saying it is, just that it can, especially depending on who and why they did it in the first place)

I thought I frames were the most accurate frames

they are, in the sense that unlike p frames they contain the whole image and so can't be misrepresented (as I mentioned before), but at the same time it needs to represent 1080x1920=2M pixels and when you add the colour you should realize there is absolutely no way to represent all that info, so it is also the most compressed part of the film (do your homework), the reason it is important is that the i frame won't be a good example for DNR because i frames must have less noise or they need to represent a lot more pixels. I am not saying it is wrong to screen capture i frames (after all if a player has issues -like an under performing HTPC used for the screen caps- that messed up the p frame then hat p frame would be even less representative) but if someone is using it for DNR then it is not at all representative of what you will see, it will be the most DNRed part of the film and it will be done by the encoder.

[4K2K]

Originally Posted by Anthony P

d'uh, so you think because someone posts useless screen shots it means they can't be useless? what good is " a rough idea " when that rough idea can be 100% wrong (not saying it is, just that it can, especially depending on who and why they did it in the first place)

Do you think the screenshots posted on the blu-ray.com site reviews section are useless? I don't think so. I believe they do give you an idea of what something is like (I'm not just talking about picture quality, but also roughly what a movie/documentary etc is like). You can also read the reviews and look at trailers too. If you don't think so, that's fine, you don't need to look at that section. Also, although the screencaps are jpegs on this site, they are quite big (>1mb for the ones I saw) so do give an idea (I know that's only possibly about a 3rd or so of an uncompressed frame). The screenshots by the people on the avs site are usually uncompressed so that should give an idea too. If the people's faces look waxy (Scary Movie? Patton? Ultraviolet?), that will also let you know about that and may influence your decision to buy. I know screenshots don't show motion, but if every screenshot of a film shows people with waxy faces, it will give you some idea of what it's like (and you can look at the trailers and read reviews as well)

[Joe Cain]

This is helpful! Thanks for the contributions made so far.

One question: what do the "I," "B," and "P" stand for?

And another: I understand that color rendering is entirely dependent on the display and that there's no way to account for calibration or its lack, but is there anything fundamentally incorrect about the color data captured? I mean, regardless of the original captor's hardware or calibration, if I'm viewing the capture on a properly calibrated display, is what I'm seeing accurate? And does compression skew hue?

[dadkins]

The data from a computer capture is as clear as the drive and software can decode it.
The compression used for the storage of the image *may* skew hue, sharpness, and clarity - depending on compression used and quality of that compression.

[4K2K]

And someone said that, for the screenshots on the blu-ray.com site, the RGB levels are converted to PC RGB levels (0-255?) so that black bars on blu-ray.com screenshots will be rgb 0,0,0 rather than the 16,16,16 that they will usually be on the disc itself.

[4K2K]

Originally Posted by Anthony P

they are, in the sense that unlike p frames they contain the whole image and so can't be misrepresented (as I mentioned before), but at the same time it needs to represent 1080x1920=2M pixels and when you add the colour you should realize there is absolutely no way to represent all that info, so it is also the most compressed part of the film (do your homework), the reason it is important is that the i frame won't be a good example for DNR because i frames must have less noise or they need to represent a lot more pixels. I am not saying it is wrong to screen capture i frames (after all if a player has issues -like an under performing HTPC used for the screen caps- that messed up the p frame then hat p frame would be even less representative) but if someone is using it for DNR then it is not at all representative of what you will see, it will be the most DNRed part of the film and it will be done by the encoder.

What I'm saying is that, because they human eye can see over 60fps (with 120hz and 240hz HDTVs available to prove this, as I said above), we shouldn't have to have I frames that are bad and then they get a bit 'better' if there are P or B frames later (and the P/B frames aren't as representative of the actual movie frame as an I frame is). If we can see higher than 120fps we shouldn't have to wait for 1/12th of a second for a good image (2 x 1/24th of a second frames). The bitrates used should be high enough that every frame is good, even on 60hz documentaries/concerts etc. so we minimize artefacts and have as high a quality picture throughout.

[Deciazulado]

If a direct view LCD DVI computer monitor is set to D65 and has sRGB primaries and gamma (2.2) (same as HDTV standard) and the capture was done correctly for RGB it should show you what it looks like. Probably better that most other displays when the image is watched at actual size (no resizing).

[Anthony P]

One question: what do the "I," "B," and "P" stand for?

I-frame = Intraframe
P-frame = Predicted frame
B-frame = Bi-directional frame
assuming you meant more then just the words,

I frame = whole frame (i.e. like the pictures that make up the film

a P frame and B frame are almost the same thing, a P frame uses the previous frame and only tells the video processor what to change, to get the new one, and B frame will also have info on the next one.

for example let's say we have the three frames of a film of a guy walking F1=upright, F2 = leg bends to take step and F3 is the guy after the step

F1

|
|

F2

|
|>

F3

|
/\



if F2 is an I frame it would have the whole image

|
|>

or
1.1=
2.1 = |
3.1 = |
3.2 = >


if F2 is a P-frame it will only have
add > to third row of F1
i.e.
3.2 = >

if F2 is a B frame then
add > to third row of F1; replace |> by /\ for row 3 for F3

i.e.
3.2 = >

3.1=/
3.2=\

so your decoder = gets F1

|
|

then it uses the info in the P or B frame to make F2
but it can make a mistake and get

|
| >

or


|
>


which will cause drift and jitter

that is why you need i frames every so often (to correct the decompressing artefacts caused by miscalculations)

but (to make it simple and go back to the initial discussion

F2 as a P frame would only need info on that one element (right leg) while if it is an I frame it would need 4. So a lot more info would be needed to have an I frame then a P (or B) frame, and the encoder will decide to use I, P or B depending on what is allowed and what is most compressible. That is why I frames end up being extremely compressed and why videos use P and B frames to get reasonable BW.

now why P and B? P are good for play but B is needed for stuff like reverse (ergo the name bi-directional)

[Anthony P]

What I'm saying is that, because they human eye can see over 60fps (with 120hz and 240hz HDTVs available to prove this, as I said above), we shouldn't have to have I frames that are bad and then they get a bit 'better' if there are P or B frames later (and the P/B frames aren't as representative of the actual movie frame as an I frame is). If we can see higher than 120fps we shouldn't have to wait for 1/12th of a second for a good image (2 x 1/24th of a second frames). The bitrates used should be high enough that every frame is good, even on 60hz documentaries/concerts etc. so we minimize artefacts and have as high a quality picture throughout.

I understand what you are saying, you are jut completely wrong
1) 120hz TVs are extremely new
2) 120Hz TVs are not as much about a higher refresh is needed but for other reasons (like 3D ready and removing 3:2 pool down)
3) yes if the refresh rate is too slow it causes issues, but that is a refresh issue and the tech not your eyes, for example CRTs had one electron beam that traveled through each of the 480 lines, the reason it was interlaced and needed such high refresh wis that when the electrons hit the phosphor on the screen it starts to glow and almost immediately starts to fade if that phosphor does not get an electron jolt fast enough then it would have dimmed enough and long enough to start getting sensed.

4) if you where even remotely right then film could never exist. Think about it, you have a film strip and on it there are different pics, there are 24 pics per second, for each of those pics for a very short period they are on screen (when the film is justr right) and there is either black as the film moves from one pic to the next or (assuming not) there is an image moving from one frame to the other.

5) in order to do I frames right you would need roughly 50mb per frame and if you can do I frames right then there is no need for p or b frames (i.e. you can't have 50mb for 1/24 of a second and then be limited to 10mb for the next one (i.e. all the frames should be able to go to 50 and then why bother adding the possibility of errors or limit play back)