xvColor and AX2000

HDR-AX2000 / HXR-NX5 (2010).
Arkady Bolotin
Posts: 60
Joined: 24 May 2010 16:46
Location: Beersheba, Israel

Re: xvColor and AX2000

Post by Arkady Bolotin »

Stephan wrote:... it looks to me like you simply increased color saturation...
Very well, you’ve hammered the nail! It’s exactly what it is supposed to look like: this new standard – xvColor (or xvYCC) – is all about more colors: 1.8 times more reds, more greens, more blues!

Really, as we know, in HSV Color model (which stands for Hue, Saturation, and Value) saturation is the distance between the white point and a given color: the further from the white point the color is, the greater its saturation.
Saturation.JPG
Saturation.JPG (28.13 KiB) Viewed 19539 times
So, if I would decide to merely saturate my footage, I had to increase all the colors’ distance from the center (i.e. the white point).

Then again, the xvYCC color standard allows reproducing colors beyond the sRGB Color gamut, that is, further from the white point! Therefore, the record shot in the xvYCC would be perceived like someone artificially increased the footage’s color saturation!
sRGB gamut.JPG
sRGB gamut.JPG (30.59 KiB) Viewed 19539 times
Therewith, there’s a simple clue showing whether the footage was saturated or not. This is the reddishness of gray patterns and human skin colors in the image. If saturation is applied they all would be strongly tinged with red. But look closely at the last video I posted: all the grey colors are truthfully grey, you cannot find any color shifting. It’s just more colors, that’s it!
Arkady Bolotin
Posts: 60
Joined: 24 May 2010 16:46
Location: Beersheba, Israel

Re: xvColor and AX2000

Post by Arkady Bolotin »

steve wrote:The 16-235 range restriction is a legacy of composite video when used for broadcasting to prevent over modulation. Since digital video and editing has been in general use, there has been the option to spread the (albeit limited) gamut over the full 8 bit (0-255) range. The removal of this unnecessary reduction of colour resolution has helped to mitigate the limitations of 8 bit video digitisation which is exacerbated when adjustments such as gamma and knee adjustments are made post digitisation.
Increasing the range of colours will put things back to the 16-235 quality in terms of smooth gradations. It's one thing comparing colour wheels side by side, but debateable whether the additional colour depth will be noticeable in prectice. For instance, Blu-ray does not support extended gamut encoding and I don't hear complaints about the colours in commercial discs. The loss of greyscale smoothness can however be detected after some compression artifacts have taken their toll where the footage is originated from 8 bit digitisation.

Steve
Hi Steve,

Sorry for the delay in answering your post.

Much to my shame I do not exactly understand the point of your argument here: either you call into question the whole principle behind the xvYCC standard, or you just challenge its particular drawback as the limited (in comparison with the full dynamic 0-256 range) colour gradation.

In either case my respond could be very partial and incomplete.

The xvYCC colour space standard (IEC 61966-2-4) was adopted in January 2006; since then it has been greatly endorsed by the Japan Electronics and Information Technology Industries Association (JEITA). As I assume those gentlemen from JEITA know what they were about to approve formally and they did their homework properly.

As to the limited colour gradation, I think we are talking about two different things: how much colour we have and how many gradations between colours we get in the picture.

The amount of colours is extended in the xvYCC colour model (that was an idea of it), the quantity of gradation (gray scale) is the consequence of the base used for signal quantization. With the xvYCC you can use 8-bit quantization as much as higher base such as 10-bit or more. Apropos, I personally used for rendering my xvYCC video 32-bit quantization.
User avatar
Stephan
Site Admin
Posts: 592
Joined: 20 Mar 2010 18:51
Location: Paris, France

Re: xvColor and AX2000

Post by Stephan »

Good discussion! So, here's for you my complete thought process before I posted that short comment.
Arkady Bolotin wrote:Then again, the xvYCC color standard allows reproducing colors beyond the sRGB Color gamut, that is, further from the white point! Therefore, the record shot in the xvYCC would be perceived like someone artificially increased the footage’s color saturation!
True, but only for those colors which are illegal in sRGB - they would ordinarily be clipped, and with xvYCC the saturated reds are redder, the saturated yellows yellower, and so on. As jbeale pointed out, the color properties of sRGB-legal colors are unchanged. Again, from BT.709 to xvYCC, only the saturated colors are even more saturated whereas the pale (unsaturated) colors do not change.
Arkady Bolotin wrote:Therewith, there’s a simple clue showing whether the footage was saturated or not. This is the reddishness of gray patterns and human skin colors in the image. If saturation is applied they all would be strongly tinged with red. But look closely at the last video I posted: all the grey colors are truthfully grey, you cannot find any color shifting. It’s just more colors, that’s it!
Indeed, I pay extreme attention to skin tones. So my first reflex was to check the skin of that woman on the far right at the end of your video. The skin tone is good, so I thought: excellent, these's more red in flowers and more green in trees, and yet the skin tone is preserved. Mission accomplished! But the grounds do look a bit yellowish/reddish sometimes.

Then, in a second step, I wondered: what we really need is an A/B comparison, because without a reference it's difficult to judge what shifts, and what shifts not. So I looked again at your side-by-side comparison (first post, Color test.jpg), and there, I'm sorry, but everything shifts: bluer sky, redder flowers, the stone/sand ground and the house wall turn yellow/red. The ugly face of too much color saturation across the whole picture.

Maybe it's linked to how you export the video and the still image, how Vimeo processes the video, how the browser displays the video and the still image... It's a complex chain of components, and we control barely nothing really. We need to make sure that your export processes map sRGB to 16-240 (or is that 235) and use the extra values for xvYCC, and so do Vimeo and all the software components. Heck, what do we know... I truly have no idea whether all that workflow can support xvYCC. Probably not: computer monitors and operating systems and software components have been set to the sRGB color space for years, and if you need to change that I don't think it can be transparent.

I don't know, this is confusing.
steve
Posts: 74
Joined: 28 May 2010 10:04
Location: UK

Re: xvColor and AX2000

Post by steve »

Arkady Bolotin wrote:
As to the limited colour gradation, I think we are talking about two different things: how much colour we have and how many gradations between colours we get in the picture.

The amount of colours is extended in the xvYCC colour model (that was an idea of it), the quantity of gradation (gray scale) is the consequence of the base used for signal quantization. With the xvYCC you can use 8-bit quantization as much as higher base such as 10-bit or more. Apropos, I personally used for rendering my xvYCC video 32-bit quantization.
The prospect of increasing the gamut with no penalties should be of interest. However my point was that for over 10 years, many have been applying the colourspace over the full 0-255 range which some editors offered as an option, especially where the output was never intended for broadcast use. This improved the smoothness of colour gradients particularly where 8bit colourspaces were used, i.e. mainly on consumer cameras.

Today, the majority of consumer cameras (and their 'professional' derivatives) still use 8 bit. Even the XDCAM EX cameras only give their full 10 bit 4.2.2. video out via SDI or HDMI. I cannot find any references to xvYCC, x.v.Color (or IEC61966-2-4 as it should be called) to any professional kit, so it would seem to be more of a consumer 'feature' available on TVs, camcorders and some PC video cards. As I said before, broadcast TV doesn't seem to use it nor does Blu-ray, and I don't see any complaints about he depth of colour on those media.

Steve
jbeale
Posts: 2
Joined: 22 Jul 2010 16:57
Location: San Francisco, CA

Re: xvColor and AX2000

Post by jbeale »

steve wrote: As I said before, broadcast TV doesn't seem to use it nor does Blu-ray, and I don't see any complaints about he depth of colour on those media.
A reasonable observation. However, the question is what are you comparing the traditional media to. When I took a video of a dark red rose and looked at the image in playback mode, it was perfectly recognizable as a red rose. When I held up the actual rose next to the screen, the screen image seemed a muddy orange by comparison. I believe it has to do with the limited gamut of the LCD display. Some newer displays which use laser illumination for the RGB primaries* have a significantly extended display gamut over the conventional displays. By all accounts the laser colors are quite visibly more dramatic than anything possible on a standard display, and (I presume) x.v.Color encoding can take advantage of that extended range.
*http://www.mitsubishi-tv.com/laser.html
Arkady Bolotin
Posts: 60
Joined: 24 May 2010 16:46
Location: Beersheba, Israel

Re: xvColor and AX2000

Post by Arkady Bolotin »

Again, Steve, you are talking about the gradation between colors – the number of the intermediate steps between two visually distinguished shades of a given color or two different colors. This gradation is responsible for the smoothness of color gradient.

Imagine you’re going to reproduce a transition between deep blue color of the skies’ zenith and white color at the horizon. Obviously, the greater amount of steps (shades of blue) you have the better (the smoother) will be this transition. With 8-bit quantization, you have principally 256 steps. As much as it sounds, this amount isn’t enough: due to the 256-steps-gradation you may see noticeable banding in the color transition.

That’s why many professional cameras have an ability to shoot using 4:4:2 or even 4:4:4 chrominance sampling (i.e. signal compressing) and 10 bit-quantization (1024 steps) for color and luma information. Furthermore, the boost in numbers of gradation bits leads to the greater precision in recording feathered edges, inter-frame fades, etc.

But again, that all concerns only the gradation not the amount of colors.

Increasing numbers of bits of gradation does not automatically mean extending the color space. That’s the reason you’ve never seen the reference about xvYCC color gamut reading information about gradation and chrominance sampling methods (4:2:0, 4:4:4 or even 3:1:1).

Once more, the higher gradation increases the number of available colors within the boundaries defined by the RGB or YCC color space, while xvYCC expands the available range (limits) to allow the display of colors that meet and exceed what human eyes can recognize.

Yes, it’s true that the Blu-ray movie format is limited to 8-bit 4:2:0 sampling and YCC color space. There are tons of explanations why xvYCC expanded color capability is not included into the BD-ROM specification. One of them is the issue of backwards compatibility: how to make legacy BD players run a new movie format rendered in 10 bit- or higher quantization in the xvYCC color gamut.

However, whatever the reason, it has nothing to do with human color perception. And here jbeale is completely right. Nobody complaints about color limitation in the modern Blu-ray movies by the same reason that not one person complained about movie colors in 60’s and 70’s: we follow the story and we would believe in colors even if they were out of the picture.
steve
Posts: 74
Joined: 28 May 2010 10:04
Location: UK

Re: xvColor and AX2000

Post by steve »

Just to mention once again: In the DV days, say 10 years ago, when I used Premiere, an option of 'NTSC compliant' and full 8 bit encoding was available in the output rendering menus. This allowed projects rendered for non-broadcast use (probably most of them) to use the full 8 bit range and not be constrained by the need to stay between 16 & 235 values. So we have the standard gamut spread over the maximum 256 values. I believe that most non-broadcast video rendered since then has taken advantage of the larger range.
If the gamut is then extended to that offered by xvYCC, you are using the 256 values to represent a greater range of luminance, Pr and Pb. The limitations of the 8 bit channel will become even more visible on certain types of image content. Whether this will be more or less obvious than the impact of wider gamut is unknown to me at least.
My reference to professional camera specs. is that so far I have not found xvYCC offered on the specs of any yet, although the product lifecycles of professional equipment are usually longer than those targeted at the more impulse-driven consumer market.

Steve
Arkady Bolotin
Posts: 60
Joined: 24 May 2010 16:46
Location: Beersheba, Israel

Re: xvColor and AX2000

Post by Arkady Bolotin »

Steve, once again, the gradation between colors (bits of quantization) and the amount of colors are absolutely independent parameters, one does not necessarily define another.

Let me take an extreme example: imagine only two colors in the picture – black and white (actually, only one color -white). With 8-bit quantization, I will have only 256 “colors” (shades of gray) to depict; with 10-bit quantization, I will have already 1024 “colors”, and so on.

Thus, the limitation in the encoding range (bits of quantization) would lead to color banding, uneven tonal transitions, and reduced contrast ratio, but it has nothing to do with the amount of true colors I have got in the image. All this may happen in the picture with just two colors, tree, and whatever you say.

Whether the 8-bit quantization would be more evident in the xvCC color space, I doubt this. I think opposite is correct: due to enlarged number of colors, the color gradient might be more subtle.

As to why xvCC isn’t included in the professional cameras’ specifications, you’re mistaken. And the point in the case is the new Sony NXCAM HXR-NX5U professional camcorder.
Arkady Bolotin
Posts: 60
Joined: 24 May 2010 16:46
Location: Beersheba, Israel

Re: xvColor and AX2000

Post by Arkady Bolotin »

Stephan wrote:...only for those colors which are illegal in sRGB - they would ordinarily be clipped, and with xvYCC the saturated reds are redder, the saturated yellows yellower, and so on. As jbeale pointed out, the color properties of sRGB-legal colors are unchanged. Again, from BT.709 to xvYCC, only the saturated colors are even more saturated whereas the pale (unsaturated) colors do not change...
I’ve just found out how to show this.

Please observe the screenshots of the vectorscope monitoring the video I recorded with and without xvYCC.

You can clearly see that those images reordered in xvYCC have the chroma values exceeded 100% of the broadcast-legal saturation for red (R), magenta (Mg), and cyan (Cy).
Attachments
Vectorscope of a crossroad shot in sRGB
Vectorscope of a crossroad shot in sRGB
Vectorscope of the building (sRGB).jpg (46.73 KiB) Viewed 19505 times
Vectorscope of the same crossroad shot in xvYCC
Vectorscope of the same crossroad shot in xvYCC
Vectorscope-of-the-Building.jpg (48.01 KiB) Viewed 19505 times
Vectorscope of a bush with red flowers (xvYCC)
Vectorscope of a bush with red flowers (xvYCC)
Vectorscope-of-Red-flowers.jpg (43.4 KiB) Viewed 19505 times
steve
Posts: 74
Joined: 28 May 2010 10:04
Location: UK

Re: xvColor and AX2000

Post by steve »

Arkady Bolotin wrote:As to why xvCC isn’t included in the professional cameras’ specifications, you’re mistaken. And the point in the case is the new Sony NXCAM HXR-NX5U professional camcorder.
I'm not sure that it is unless the Sony professional range is going over to consumer microphone sockets like 3.5mm. I doubt that pro users would take that very seriously. There is talk here of a Sony Professional version in a few months, just like the Z1 came out of the consumer division's FX1, so that would be a case of a professional camera inheriting a consumer feature by default. As I said before, so far I have not found xvYCC offered on the specs of any professional, (please read that as a camera developed as a professional piece of kit to meet the needs of the pro user).

Steve
Post Reply