I own a blackmagic intensity pro capture card. I'm still learning and was wondering if one could achieve a perfect uncompressed capture of some input.
This is my setup: Computer1 Nvidia output -> Computer2 Intensity pro hdmi input.
To the naked eye it looks uncompressed but when analysing pixels closely there is some noise or color differences.
Same results with both 720p@60 and 1080p@30. I tried with blackmagic media express and VirtualDub 1.10.4
I can spot those differences in a video captured with bm media express, and in the preview window of virtualdub (not even recording)
VirtualDub settings: Video -> Compression... -> "(No recompression: HDYC)"
Video -> Set custom format... -> Custom size 1920x1080 / Data format: HDYC YUV 4:2:2 interleaved (Rec. 701) (only working data format)
Does anyone have any idea about some options or else that would cause this?
+ Reply to Thread
Results 1 to 14 of 14
You're going to have to show us some examples of what you mean by "noise or color differences".
Your captures are always going to be 4:2:2 with this device, which means that colors will be 1/2 the resolution of the greyscale.
You may be viewing your caps with a rec.601 matrix rather than a rec.709 matrix. That will cause a shift in (mostly) greens and reds:
Even if you're using the correct matrix, if you're starting with RGB images you will lose accuracy by the conversion to YUV and colors will blur because of the chroma subsampling.
Thank you both for your help. Well I didn't know much about chroma subsampling until now. It's indeed the 4:2:2 subsampling that bothers me. Is there any card that will capture hdmi raw rgb input with 1080p@60 ? (4:4:4 I guess?).
I need at least yuv 4:4:4 for preview mode in virtualdub, not necessary in file capture mode.
AVCHD, BluRay, and DVD standard colorspace is usually planar YV12 (4:2:0). Rec601 is used for SD, Rec709 for HD. Video is displayed in VirtualDub as decoded Rec601 RGB24 regardless of the colorspace of the source, assuming it's a source that VirtualDub can read. As most of the NLE's and other processing software that consumers use will work with 8-bit color (which is what your monitor and TV also use), you're pretty much stuck with that unless you get more sophisticated software and hardware.
We don't know what you're capturing. If it's VHS or similar, it's usually captured as lossless YUY2 or sometimes as RGB. You're getting into an area which requires more research and the gear to handle it. If you submitted a piece of HDYC video for evaluation, many members wouldn't be able to display it properly.
Last edited by sanlyn; 19th Mar 2014 at 09:42.
Since you're recording from a computer you should try using screen capture software instead of hardware. That way you can get RGB 4:4:4, lossless.
Some of the Avermedia capture cards have HDMI input uncompressed (or losslessly compressed) capture. I don't know if they'll work in RGB.
The source is RGB (as set in the nvidia control panel) but my intensity pro card applies subsampling 4:2:2 and that's my issue. I want to display true RGB (4:4:4) under VirtualDub.
Is this project worth thousands of dollars to you?
And keep in mind that almost all final delivery formats are YUV 4:2:0.
Found this card (~300$) :http://www.timeleak.com.cn/en/article/view.php?ps_db=prod&pnid=2&ps_aid=18
*HDMI and component video: Up to 1080p/60Hz
*Maximal resolution of DVI: Up to1920*1080 60 Hz
(And by the way, dvi is fine too)
*Hardware color conversion, with output formats of RGB24, RGB32, YUY2, UYVY and I420
Capture Format : Color space: YUY2, UYVY,RGB24, RGB32, I420
Now this is a bit confusing to me. They state rgb24 as output option so that's what I'm looking for right? Although I read this on the virtualdub doc : "Guess how they produce RGB24 data when you ask for it? You guessed it, the YUV 4:2:2 data is upsampled to RGB24. This means that the extra color resolution that RGB24 gives is entirely fabricated."
So i'm wondering if this card auto applies yuv 4:2:2 conversion.. I emailed them..
I don't know what you mean by subsampling. Except for animation and other forms of compuiter-generated audio/video, all audible and visual phenomena in nature originate as analog phenomena, and all visual stimuli as seen by humans originate as RGB. This assumes you're not trying to view infra-red/infra-violate, etc., which themselves are resampled to perceptual RGB levels in order to be viewed. If there's any downgrading of original events, it occurs when analog video or film is truncated and digitally rounded to common formats like YV12, which represent less than 60% (or less) of the original data. How any particular display device gets YUV back to RGB is widely documented and standardized, even if some display devices such as most LCDs and TV's do a very sloppy job of translation unless they're properly adjusted (which we like to call "calibration"). If you are this concerned about display fidelity, you would by now have learned to calibrate your display devices and graphics cards with a colorimeter and specialized software designed for that purpose. Because display devices behave differently over time, most people repeat the calibration at least every 3 months. This includes selecting a PC monitor and graphics card that respond well to those calibration kits, as well as selecting a TV or other video monitor that is equipped with a minimal set of CMS (Color Management System) controls to properly set grayscale and gamma points.
All colorspace conversions entail compromise and error of some kind, if for no other reason than as a result of digital rounding during interpolation. There is no getting around this. There are proper methods for doing it that can minimize loss. There are inferior and superior methods for capturing/recording different digital and analog sources. There is no "perfect" way of doing any of this, there are just bad ways and better ways.
Don't know how I got into all this detail. I suppose describing a source as "RGB" doesn't tell us that much unless what you're capturing or recording is some form of game play or other computer generated graphics.
Last edited by sanlyn; 19th Mar 2014 at 09:42.
unless you are creating RGB444 from scratch, i would ask, what could you possible have that is actually RGB444 that you need to capture it exactly that way over hdmi ? since any input video obtained through some capture devices will originate from YUV420 color sampling format, and even that will be internally processed to YUV422, which is what you will ultimately want to capture as your final destination file in this goal of yours. anyway.
like everyone else said, tipicle capture cards internally process the incoming source video to a finished uncompressed (raw) though sampled YUV 4:2:2 color format. from there, you have the choice to capture that as either:
so, the possibilities may include any of the following over composite/svideo/component/hdmi/sdi/dvi input connections to the capture device ...
input YUV420 -> process -> output YUV422
input YUV420 -> process -> output YUV422->RGB444*
input YUV420 -> process -> output YUV422->YUV420*
where in the case of "process" the capture device internally processes the video in 8,9,10, or 12bit. higher bit depth gives better quality.
* where the capture device is selected to produce rgb444, it will yuv422->rgb444.avi
or, if selected to produce yuv420, it will yuv422->yuv420.avi
i don't know any consumer cards that take a YUV422 as an input source, since vhs, laserdisc, dvd players, dvr recorders, satalite receivers, etc., etc., all output yuv420 only.
VirtualDub's documentation is super old, so that line is referring to analog capture cards. The card may be producing "fake" RGB, or maybe not. Customer testimonials for the Timeleak cards are hard to come by. Hopefully the company is able to respond with accurate info.
NTSC VHS isn't 4:2:0 either. It's an analog signal with roughly 2:0.1:0.1 resolution on a "4" scale.
In any case, it doesn't matter if the OP is capturing RGB from a computer. His source could well be 4:4:4.