Printing stills captured from HDV footage

27 replies [Last post]
David L Lewis
Offline
Joined: Jan 11 2006

Just having a play with Premiere pro and Photoshop, (CS3)

Ive captured a still from HDV footage, simply by exporting a frame and premiere pro creates a BMP file that is 4.577MB and has a frame size of 1440 by 1080.

Now Ive opened the file by clicking on the file under my computer/my pictures and adobe photoshop opens up and the captured picture is displayed but it seems a bit squashed . presumably because the picture is 1440 pixels wide rather than 1920 (maybe something to do with square pixels instead of oblong ones????)

How do I get the picture to be the correct width. do I need to alter a setting in premiere pro when I create the file or in Photoshop when I'm viewing it.

Also can premiere pro create pictures in other formats i.e JPG or Giff

David L Lewis

Hello I'm in Mensa, Is there anything you would like me to explain to you?

ChrisBitz
Offline
Joined: Jun 8 1999

I'm not sure in adobe cs3, but if you search help for "pixel aspect ratio", you'll find where the menu item to correct it is...

or you can resize the picture to 1920 wide, and uncheck the "contrain proportions" tickbox.

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

Very few tv pictures have square pixels, this is something you always have to bear in mind when importing or exporting stills.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

David L Lewis
Offline
Joined: Jan 11 2006

Ok managed to answer my own question by having a play with photoshop and altering the image size from 1440 width to 1920 ( and deselecting constrain proportons)which seems to have done the trick and produced a print of the right dimensions. But it would be nice to know if i could get it right at the capture stage

Sorry chris just seen your reply

David L Lewis

Hello I'm in Mensa, Is there anything you would like me to explain to you?

ChrisBitz
Offline
Joined: Jun 8 1999

it's not possible to "get it right at the capture stage" Non-square pixels are just the way that Television works.

you may be able to stretch the still frame when you export it from your editing program, but if it comes from a video camera, it will be made of non-square pixels.

David L Lewis
Offline
Joined: Jan 11 2006

seems a bit odd that its not possible to change things at the capture stage.

a fairly basic program like Pinnacle studio lets you "grab" a frame and that saves itself in the right size. Dont know how it does it but you just grab your frame and everything is in the right proportions.

David L Lewis

Hello I'm in Mensa, Is there anything you would like me to explain to you?

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

You're using a video camera to record video. It does that correctly. The pixels aren't square, they're not supposed to be. For import/export of stills, some NLEs will assume that you don't know what you're doing, and adjust aspect ratios automatically, these are typically the expensive ones (e.g. FCP). The cheaper ones assume you know what you're doing and let you make your own decisions. And I assume you know all about the Rec601 data levels and are doing the right thing, or do you want me to explain it to you? I've done it many times already though.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

ChrisBitz
Offline
Joined: Jun 8 1999

HA! That really made me laugh!

I was just about to post something about The Apple school of technology where "you can't change that because you wouldn't want to, ok?" but i thought it might come across as a bit rude. :-)

I can't count the number of FCP editors who've given me edited footage at 1024x576, and I've bounced it back to them saying "I want it as 720x576 - the proper size for widescreen" and then they call me all sorts of names or just look at me as if I'm an alien, or don't have a clue what I'm talking about!

Here's a bit of further info, but it's a bit techie, but skim it and there are *some* good answers in there.

http://lurkertech.com/lg/pixelaspect/

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

It's the rigid inflexibility of FCP (plus the mistakes it makes) that's put me off Macs.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

StevenBagley
Offline
Joined: Aug 14 2000
ChrisBitz wrote:
I can't count the number of FCP editors who've given me edited footage at 1024x576, and I've bounced it back to them saying "I want it as 720x576 - the proper size for widescreen" and then they call me all sorts of names or just look at me as if I'm an alien, or don't have a clue what I'm talking about!

Is it actaully 1024x576 or is quicktime actually just reporting it is 1024x576? QT lets you specify the pixel-aspect ratio and uses that when displaying the video but it is actually stored as 720x576 (it does the same for 768x576 and 4:3 too).

It's an incredibly useful feature, since it means you can correct all those videos that have the wrong aspect (which I seem to get sent!).

Steven

ChrisBitz
Offline
Joined: Jun 8 1999
StevenBagley wrote:
Is it actaully 1024x576 or is quicktime actually just reporting it is 1024x576? QT lets you specify the pixel-aspect ratio and uses that when displaying the video but it is actually stored as 720x576 (it does the same for 768x576 and 4:3 too).

Steven

I'm not sure if I understand you correctly, but in my experience, the videos I get from the FCP guys are definitely 1024x576, rather than 720 with an aspect ratio correction...

Is that what you meant?

(I'm importing it into After Effects rather than just playing it in QT)

StevenBagley
Offline
Joined: Aug 14 2000
ChrisBitz wrote:
I'm not sure if I understand you correctly, but in my experience, the videos I get from the FCP guys are definitely 1024x576, rather than 720 with an aspect ratio correction...

Hmm, I do this a lot and they are always 720x576 as you'd expect... I'm looking at a clip which was shot HDV and then downconverted in FCP to 4:3 DV (AE gets a bit weird with HDV clips in my experience and I needed it 4:3 anyway) and it is correctly being reported as 720x576 with a 1.07 pixel aspect (well the aspect is wrong but all Adobe/Apple software gets that wrong -- should be 1.094). Using AE CS3 (Mac) and FCP 6.0.3.

In fact, I've just exported a short section of a 16:9 DV clip (but untagged in the DV stream due to its providence) from FCP and imported it into AE. AE happily reports the clip as 720x576 (DV codec)...

Steven

Dave R Smith
Offline
Joined: May 10 2005

I don't know if you're convinced yet David, there is no 'right output' - there is a variant (square ratio pixel) which you may use for pc monitor or web use and a widescreen (1.422) pixel ratio for use in TV and TV edit systems.
In photoshop, if you have your 720*576 image, by default it will display with square pixels, but there is an option 'image/pixel aspect ratio' to view various other ratios.
Don't forget if your NLE exports SD as 720*576 (1.422 ratio) it is an exact pixel for pixel copy.
If you select say 1024*576 (square ratio) your NLE has to juggle the image to respread horizontally, so the result is not exactly the same as the source - so if you later suck it back in to your NLE you will suffer a 2nd reshuffling of pixels which we normally want to avoid in a world of digital copying where we strive for zero loss.

H T H ;)

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

Let's back-track a bit.

All SD video in the 625 form, is 720x576. There is no 1024 version. Video does not have square pixels. It's in all the standards. The only juggling of pixel shapes that any NLE ever has to do is when importing or exporting stills. The video always has the same pixel aspect ratio, irrespective of the image aspect ratio. When you export to anything other than a television device, it's not video any more, it's something else and we haven't got standards or terms for it yet. So exporting for YouTube for example, isn't a video standard, it's something else.

HD works the same way, the images are 16x9 but the pixel dimensions will be 1920x1080 or 144x1080, or 1280x720 or 960x720, depending on the source.

The NLE should always display the video in the correct aspect ratio (4x3 or 16x9) and might display the image dimensions assuming square pixels, but those are only notional (and wrong in the case of Apple/FCP), they don't relate to the video image itself.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

Chris.
Offline
Joined: Nov 5 2000

Have just read the lurker link. It mentions that HD 1080i and 720p are square pixel (I hadn't realised that, but I've got no HD equipment at all yet so haven't really read up on it too much).

The reason I didn't think they were square was that after glancing at specs of HDTV sets a lot mention a resolution of 1360 x 768. I never really understood how that tied into the HD resolutions and had just assumed it had something to do with square pixels versus rectangular pixels.

This will have been covered in a previous thread, grateful if someone can point one out if so - what's the deal with 1360 x 768, where does that resolution come into the HD equation?

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

Chris, the HD standards, as enshrined in the SMPTE and ITU specs, are all square-pixel. So we have an image shape with an aspect ratio of 16:9, with two possible sets of dimensions, 1920x1080 and 1280x720. However, much of the acquisition is at 1440x1080, or 960x720 in the case of Varicam, purely for reasons of economy (bits cost money). So, although square pixels are the norm in the standards, much of the production ignores that.

Now, the displays.

Flat panel displays came out of the computer industry, not the consumer tv section, so they tend to have dimensions (in pixels) that make sense in the VGA hierarchy. The first panels we saw around 15 years ago (IIRC) were 640x480 (VGA, 4:3), then came 1024x768 (XGA, 4:3) and the first wide-screen big panels were 853x480 (wide VGA, 16:9). When HDTV markets started to get significant, mid 1990s, we started getting bigger panels e.g. 1366x768 (WXGA, 16:9). I have a Sony 20" WFM that's 1680x1050 (WSXGA+, 8:5) which correctly shows 16:9 pictures with top/bottom black bars, and my laptop displays are both 1920x1200 (WUXGA, 8:5), deliberately chosen so that I can see 1920x1080 pixel-mapped.

Note that none of these display sizes have anything to do with video, they're all computer display types, but they all have square pixels. There's one fly in the ointment though, there's an early panel type that has a form of interlace, dimensions 1024x1024 but 16:9 aspect ratio. It was short-lived and unpopular, but cheap, the market has moved on.

Manufacturers will make whatever they can sell. In the consumer entertainment part of the industry, that means tv sets, so the aspect ratio has to be 16:9, or has to accommodate it correctly. But they still have VGA or DVI connectors so that you can use them as computer displays as well. The 8:5 displays are all in the CGA hierarchy (CGA 320x200, WSXGA+ 1680x1050, WUXGA 1920x1200, WQUXGA 3820x2400), not the VGA set (VGA 640x480, SVGA 800x600, XGA 1024x768, SXGA+ 1400x1050, QXGA 2048x1536, UXGA 1600x1200, QUXGA 3200x2400), and occasionally you'll see some odd 5:4 displays (SXGA 1280x1024, QSXGA 2460x2048).

Displays that are intended solely for television are all 16:9 and tend not to have VGA connectors, just DVI and/or HDMI.

The reason why displays don't have the same number of pixels as the transmission format is to do with cost and common sense: the more pixels you have on-screen the higher the cost, that's pretty obvious. But, it's also down to the size of the display: there's no point having a 1920x1080 display of it's only 3" diagonal for a mobile phone, so we see whole rafts of different sizes and pixel dimensions, depending on the intended useage. 1366x768 is a common size for panels up to about 37", where the pixels are small enough when viewed from about 3m, but they look coarse when fabricated at 42" and above.

Does that help?

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

PaulD
Offline
Joined: Aug 31 2002
Alan Roberts wrote:
It's the rigid inflexibility of FCP (plus the mistakes it makes) that's put me off Macs.

Hi
Not knowing how to use FCP 'flexibly', and not knowing how to drive OS X, are an even bigger 'put off'... :(
Whilst the Apple way of doing things has many imperfectly implemented fundamentals, particularly in the antiquated intricacies of QuickTime, ALL other affordable (and much that isn't) A/V software is similarly, or even more, compromised. IMO... ;)

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

OK Paul, get FCP to do just this simple task and I'll retract.

Extract a single frame from a video timeline, preserving the sampling levels, in a Windows BMP file. One action. So, black must be at 16, white at 235. It was my failure to get FCP to do this simple task that pushed me to Edius, which does it correctly. I couldn't find any way to convince FCP that I wanted a still video frame and not a graphic file conforming to sRGB.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

PaulD
Offline
Joined: Aug 31 2002

Hi
If your intended output is a RGB still, ANY export from YUV movie colourspace is going to involve a cross-conversion. Instead do the YUV>RGB transcode inside FCP, with FCP in RGB colourspace mode, with filtering (Proc Amp etc) to create the correct levels. Also there are QT codecs that handle the YUV<>RGB conversion far better than aquisition codecs like DV or HDV.

Or if you have a YUV timeline use Photoshop with colour management turned off to to manage the creation of the BMP. And set the pixel aspect ratio to your requirements before Exporting...

From my point of view I couldn't edit in an application like Edius that didn't/doesn't? maintain an independent exportable alpha-channel in the Timeline.
If its the best tool for your test card creation that's fine ;)

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

In the case in question, import had been from HDSDI. I wanted to extract a single frame for noise analysis. I found that FCP insisted on translating the video to sRGB levels (and that the interpolation to do so is very crude), which distorted things too much. I now use a HDSDI direct capture of the data stream and extract what I need using my own software, which I know to be accurate. Needing filtering to recreate the correct sampling levels rather destroys the whole point of the exercise, which is to find out exactly what the camera emits as a signal.

For SD or HDV, Edius does what I need, because in that case I'm more interested in what signals the user has, rather than the theoretical values I need for camera analysis. Plus, the only translation involved is the YUV to RGB matrix, the ITU709 decoder, and since that's always in the signal route, it's relevant anyway.

The conversion from YUV to RGB and vice versa is a bugbear in all this, and that's why I now use the HDSDI data-stream directly for analysis of broadcast cameras, I can get at the Y signal unprocessed.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

PaulD
Offline
Joined: Aug 31 2002
Alan Roberts wrote:
I found that FCP insisted on stretching the video to fill sRGB levels, which distorted things too much.

Hi
That's capture codec dependent, and Apple's Shake or Adobe's After Effects allows more control of the video-level/gamma-conversion settings of some QT codecs - it has to, as AE is an RGB application.

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

Ah, yes, that was the other reason; I'd forgotten Apple's insistence on applying a gamma, that makes it completely useless to me.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

PaulD
Offline
Joined: Aug 31 2002

Hi
To you, for computational analysis, yes ;)
For the rest of us, viewing video in the real world, gamma encoding/decoding - controllably, as Apple software allows - is essental:

http://en.wikipedia.org/wiki/Gamma_correction

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

The problem is that the initial gamma-correction in the camera is "correct" for the tv display (that's what it's there for), and the techology for tv displays these days is the same as for the computers. So the overall transfer characteristic is about the same, whether on tv or on the computer. In many cameras, it's not possible to turn gamma-correction off, and it's highly undesirable to shoot that way anyway for a whole raft of reasons. So, why is the video gamma characteristic any business of the NLE? Surely it should be "pictures out=pictures in", why have a systematic gamma correction at all, in the NLE? Only in MACs does this happen. Of course, manipulation of gamma for effect is essential, that's called grading, but Apple insist on a two-part gamma processing, which wrecks the simple relationship between input and output. Disastrous for the science I have to do.

I spent many happy years measuring displays and cameras professionally, and I still can't see why Apple insist on using an intermediate gamma for video.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

PaulD
Offline
Joined: Aug 31 2002
Alan Roberts wrote:
I still can't see why Apple insist on using an intermediate gamma for video.

Hi
I don't think they do these days...
Quote from the FCP manual - for viewing/previewing of the source files non-destructively:
"Y CB CR media has an implicit gamma of 2.2. You cannot adjust the gamma of Y CB CR clips.
In RGB Final Cut Pro supports gamma adjustment for the following file formats:
Still images: JPEG, PNG, TIFF, SGI, PlanarRGB, MacPaint, and layered or flattened Photoshop (PSD) files.
QuickTime movie files: Movie files using the None or Animation codec."

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

Good, that's a step in the right direction.

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.

Chris.
Offline
Joined: Nov 5 2000

Just wanted to nip back into this thread to thank you Alan for the excellent answer to my question on page 2, much appreciated.

Alan Roberts
Alan Roberts's picture
Offline
Joined: May 3 1999

My pleasure Chris, lurking about and explaining is what I do :D

Get my test cards document, and cards for 625, 525, 720 and 1080. Thanks to Gavin Gration for hosting them.
Camera settings documents are held by Daniel Browning and at the EBU
My book, 'Circles of Confusion' is available here.
Also EBU Tech.3335 tells how to test cameras, and R.118 tells how to use the results.