V1 Progressive Sharpness?

[Robert Ducon]

We all know that often when one deinterlaces footage in many or most post programs, half the lines are thrown away and the remaining lines doubled. I've been a following the XH A1 and apparently when shooting in F format (24F or 30F), users have alluded to this occurring at some level because the sharpness decreases.

My question to V1 owners is, since the V1 is a TRUE progressive camera as Spotted Owl has said, does it avert any line doubling voodoo of any sort? Is this the sharpest progressive camera that shoots HDV that we know of? Or is the image a wee bit softer? I'd like to know.

A 60i/50i frame grab and then a 30P/25P frame grab from a few seconds later on a detailed area would be great for comparison.

[Douglas Spotted Eagle]

First, please get my name right? :-)
Second, this camera doesn't deinterlace, because it's true progressive. There is no line doubling (nor is there any in the canon, either)
There is no way to give you a frame grab of 60i with 30p "a few seconds later" because there is no way to switch the camera modes that fast.
You can have similar footage of the a similar event, or you can have stills of the exact same.
Sharpest camera? I'm not willing to go that far, and say it's sharper or less sharp than any other HDV camcorder, as the XLH1 is very sharp, and all of the cameras have sharpening tools. The question isn't if it's the sharpest, but does it at default settings, have a sharper image.

I've not tested the XLH1 at default against the V1u with regard to sharpness.

[Dave F. Nelson]

>>My question to V1 owners is, since the V1 is a TRUE progressive camera as Spotted Eagle has said, does it avert any line doubling voodoo of any sort?<<

No.

>>Is this the sharpest progressive camera that shoots HDV that we know of?<<

Possibly, but the jury is still out. Let's see what the reviewers have to say. Adam Wilt, in DV Magazine, February 2007, on page 48 said the following:

"Overall, the raw resolution of the prototype V1 surpasses that of either the Z1 or the Panasonic HVX200, and may even challenge Canon's crisp XL-H1. I'm (Adam Wilt) eager to test the final, shipping version."

>>Or is the image a wee bit softer? I'd like to know.<<

See above.

Hope this helps, Dave.

[Steve Mullen]

Originally Posted by Robert Ducon

... users have alluded to this occurring at some level because the sharpness decreases.

1) With Sharpness set a NORMAL, my posts here show no difference between I and P.

2) With Sharpness set a 5, my posts here show 24p to be softer than 60i.

So the solution is simple -- adjust Sharpness to taste. It's in your hands how the image looks.

[Bob Grant]

Spot and Steve,
I recently did some very crude tests on our V1P and I noticed something interesting. I did post this before but the thread got locked and I think it's just got lost amongst a whole lot of other stuff.

Anyway I found no visual difference between 25p and 50i, yes this is an updated camera! But what I did notice was an increase in noise, as expected. I was shooting an almost totally black subject, a very large, black Pelican flight case. But the noise simply wasn't there on the flat areas of the case. It was only there on the edges. No not crawling things, just strong video noise.

Now the only explaination I have for this effect is the camera is using some form of DNR and that's certainly a valid technique. However I'm wondering if what some are seeing and others are not relates to how this affects resolution at lower light levels. Obviously also given the difference in noise levels inherent in 'p' Vs 'i' it's going to be more obvious in 'p' than 'i'.

Perhaps what'd further confuse things is fine moving detail in low light and DNR. The DNR might wipe out the fine detail mistaking it for noise. I'm more than happy to run some more tests to further investigate if anyone thinks there's any merit in my ideas.

I'm far from certain that anything's 'broken', just trying to clear up some of the FUD.

[Edit] Should have mentioned I was looking at the component output from the camera on a Dell 24" LCD. Not the recorded signal.

[Douglas Spotted Eagle]

Very good point, Bob. There is a DNR system in the DSP of course, and with certain settings in the gamma setups coupled with low light, I've noticed that the DNR can get aggressive.

[Steve Mullen]

Originally Posted by Bob Grant

But what I did notice was an increase in noise, as expected.

There should be no difference in noise -- or sensitivity -- with Sony's system.

Every 1/50th or 1/60th second, a 960x1080 FRAME is input to the EIP which creates one FRAME of 1080-lines (1920-pixels wide). It makes no difference if I or P.


I-OUTPUT: every 1/50th or 1/60th second, the odd-lines are output as an odd field of 1440x540. Every alternate 1/50th or 1/60th second, the even-lines are output as an even field of 1440x540.

When viewed, either in our eyes, or in the HDTV's deinterlacer, the two fields are combined. NOTE THAT THE VIEWED "I" FRAME COMES FROM TWO (2) CAPTURED 960x1080 FRAMES!

The camera's anti-alias filter is set to optimize the FRAME when displayed. It is the anti-alias filter that determines how much stepping is seen on edges of diagonals.


P-OUTPUT: every 1/50th or 1/60th second, the odd-lines are output as an odd field of 1440x540. Every alternate 1/50th or 1/60th second, the even-lines are output as an even field of 1440x540. HOWEVER, THE *SAME* CAPTURED 960x1080 FRAME IS USED FOR BOTH OUTPUT FIELDS. (The other captured 960x1080 frame is discarded.)

When viewed, either in our eyes or in the HDTV's deinterlacer -- the two fields are combined. NOTE THAT THE VIEWED FRAME COMES FROM ONE (1) CAPTURED 960x1080 FRAME.


NOTE:
Viewed "I" comes from twice as much source information as does viewed "P." There is every reason for "I" to look better.


THE ISSUE:

Ideally, there should be two settings for the anti-aliasing filters. In "I" mode, the filter frequency can be higher because when the fields are combined, they complement each other. This helps prevent visible aliasing (stair-stepping). Keeping the filter frequency high keeps overall detail high.

In "P" mode, the filter frequency could be set lower to present aliasing (stair-stepping). This would prevent any appearance of aliasing in P. The price, however, would overall detail loss!

Remember, you can always trade-off "detail" for a "cleaner" pix. Canon has always gone for clean. Sony -- see Adam Wilt's reviews of Sony DV camcorders -- goes for greater detail which means more aliasing. Moreover, while stair-stepping is what you see on hard edges -- softer edges just get noisey. And, Sony compared to Canon has also always had noisey edges. (These have been a very common complaints.)

Obviously, the V1U seems to have the balance down very nicely. But, I don't think I would say I and P are identical. And, I don't think it's possible to be perfectly identical. I think Sharpness needs to be adjusted to your own needs in each mode. And, I can't promise Sharpness will reduced aliasing as it depends on the frequencies altered by Sharpness.)

Please note, I'm making no comment on the V1E. I'll leave that to those who have them.

[Bob Grant]

Steve,
thanks for that detailed explaination. When I get the chance I'll go back and repeat my tests on that basis with our V1P. At the time I was looking more at the aliasing non issue.

[Tony Tremble]

I wasn't going to post on this subject but I see the message still hasn't got through.

It is important for those people in 25P lands to see the issue at first hand.

1. To say the aliasing issue is a non issue is strange. In the image attached below it is very obvious that aliasing is an issue. Believe me when the tree moves in the original it looks like there is a gridded filter placed over the screen as all those edges run like crazy.

2. There is also an abundance of mosquito noise round the edges.

3. Notice also that the fine detail in the field behind the trees is obliterated by large block noise that is horrendous when in motion.

4. The oil paint effect is still present but is not so visible in this image. In images I have sent Sony it is still visible.

5. Over sharpening of edges.

These artifacts are not consistent with the very high quality of 50i.

******IMPORTANT*******
Sony are aware of ALL these issue in the 25PScan mode. These are NOT non issues to Sony. I have spoken to senior people within Sony, my dealer has spoken to Sony, Simon Wyndham emailed me and said his contacts were aware of the same issues and more. He had viewed the output on a Sony LMD23 monitor which showed "horrendous aliasing." Everyone is singing from the same hymn book. The 25P problem is real it is acknowledged and Sony are fixing it.

Thankfully it isn't a non-issue to Sony. So let us not keep inaccurately portraying the 25P issue as it doesn't seem to help anyone.

TT

The picture....

[Kristin Stewart]

Thanks Tony for keeping us informed of the Sony fix. I hope you'll have better news soon ! I called Sony France Pro regarding this problem and, as usual in France, nobody knows anything about it ! I really wish everything's gonna be OK and I don't understand why it's easier for Sony to deliver a 24P and 30P good version than a 25P one.

Kristin

[Bob Grant]

In my tests I was, at the request of another forum member looking for "staircasing" on high contrast diagonal edges. What I saw was the exact same results as I expected, the same in both 25p and 50i. As a further test we swapped the V1P for a Z1 in 50i and it was exactly the same. There will always be staircasing on diagonals at the resolution limit. This was using white card on a black background but pretty well lit. A 4 bank fluro 1 M away to be exact.

Now looking at the supplied image and blowing it up to 200% in PS I can see exactly what I saw in my tests but didn't look at in any great detail but have mentioned before.

The branches against the sky show expected staircasing. However around them is what looks like the exact same noise that I noticed in darker areas of the frame. Combine the normal aliasing (staircasing) with that noise and in motion things I agree could look really bad, I don't think it's the same as the mosquito noise that was the bane of DV though.

The dark field behind the tree has indeed become something more like a painting that a photo. Again I suspect this has something to do with dynamic noise reduction. From what Steve said before though I cannot see why it's an issue in P and not I. I was pretty certain in my tests I only noticed it in progressive and based on what was said was going to recheck it in case I had lost my marbles.

And from what I've been told Sony Australia are also investigating a problem post the first fix. What I hadn't been told is if they'd confirmed the problem or what the problem was so I wasn't going to mention it until further details emerged.

[Tony Tremble]

Originally Posted by Bob Grant

The dark field behind the tree has indeed become something more like a painting that a photo. Again I suspect this has something to do with dynamic noise reduction. From what Steve said before though I cannot see why it's an issue in P and not I. I was pretty certain in my tests I only noticed it in progressive and based on what was said was going to recheck it in case I had lost my marbles.

I agree there is always going to be some aliasing especially on high contrast edges especially in a 4:2:0 colour space......But.....in interlaced mode the stairsteps are much reduced because they are not subjected to the extreme level of sharpening.

You are absolutely correct about the dynamic DNR. In progressive mode it is over active and in changeable light one can see the detail literally is rubbed away to be replaced by blotchy block noise as the light drops. In interlace mode the detail remains even in the darks through a wide variation in light intensities.

The over active DNR and the increased sharpening creates the "oil paint" effect. The "fix" at Christmas reduced the point at which the DNR became active but in no way did make 25P acceptable.

If someone could tell me where I can get DVHSCap (Intel mac) to record short M2Ts as still frames do not show how bad the aliasing is, how that can create sparkling noise on all bright edges and to see how quickly the block noise becomes severe. The overall apparent noise is significantly higher in progressive mode as a result of these factors.

I find it hard to believe that a camera could be released in such a state and then be "fixed" and still it is not fit for purpose. Embarrassing...

TT

[Bob Grant]

Originally Posted by Tony Tremble

If someone could tell me where I can get DVHSCap (Intel mac) to record short M2Ts as still frames do not show how bad the aliasing is, how that can create sparkling noise on all bright edges and to see how quickly the block noise becomes severe. The overall apparent noise is significantly higher in progressive mode as a result of these factors.

I find it hard to believe that a camera could be released in such a state and then be "fixed" and still it is not fit for purpose. Embarrassing...

TT

I think this issue highlights the point that existing test methods are inadequate for cameras using advanced image processing. I'd bet you could run all the standard res chart tests and not see this problem.

I'll try and capture some footage using Vegas this week.

[Tony Tremble]

Originally Posted by Bob Grant

I think this issue highlights the point that existing test methods are inadequate for cameras using advanced image processing. I'd bet you could run all the standard res chart tests and not see this problem.

I'll try and capture some footage using Vegas this week.

Absolutely correct.

Check out these two frames taking care to notice the roof detail on the red brick houses in the front left and the grass in the foreground. These 2 frames were are part of the same clips separated by <20secs! ****BOTH shot in 25PScan Mode****

BTW, this does not happen in interlace mode. Interlace holds together very well.

TT

Forgot to say, light drop caused by cloud obscuring sun! No settings change between images.

[Zsolt Gordos]

Originally Posted by Tony Tremble

These 2 frames were are part of the same clips separated by <20secs! @25PScan Mode

Does this mean both images were taken with 25p? One looks as if it was interlaced (details are OK) and the other is smudged soft. I dont really get this if both are 25p