HV20: Cinemode Softness (loss of image detail) Pic

Joseph H. Moore · September 25th, 2007, 04:48 PM

Quote:

It is bad design because Cinemode is orthogonal to Av or Tv and nothing should prevent from using Cinemode (image enhancement setting) with either Av or Tv.

CINE mode is not just CINE gamma ... which is what you (and I and everyone else) wishes we had independent access to. CINE mode a combination of features glommed together into a mostly automatic consumer MODE.

Quote:

Canon either deliberately crippled the camera settings or haven't given a good thought to it at all.

Of course it is deliberately crippled ... the HV20 shares the same electronics as in a camera that costs three times as much! Sure it sucks, but it's not as if Sony et. all don't do the exact same thing. Again, that's the game you play with a sub $1k consumer device.

Euisung Lee · September 26th, 2007, 06:18 PM

Quote:

Originally Posted by Euisung Lee

Unless you are capturing live HDMI feed you always have to make a trade between low frequency detail and latitude, neither of which is recovable.

I guess I should've said high frequency not low. Anyhow...

Quote:

Originally Posted by Noah Yuan-Vogel

..im getting mtf50 numbers around 590 for cine and 670 for tv, but when sharpening is normalized the cine jumps to 650 and tv doesnt change much.

Noah, can you tell me how you normalized sharpness? Did you get 590 under cinemode with sharpness -1 and 650 with sharpness 0?

Thanks for doing the test.

Noah Yuan-Vogel · October 1st, 2007, 04:39 PM

Actually the software used to do the test adds sharpening to the images to make them (ideally) equal in terms of digital sharpening. The result is the measured resolution gets much higher for cine mode and very slightly higher for the Tv stuff that is already sharpened more than cine. So it becomes more obvious that while cine is soft, it has almost as much actual resolution as Tv, but does still appear to be overall a bit less able to resolve detail. Once i get around to analyzing the latitude tests and gamma curves, it might be able to tell me a bit more about whether the contrast has a significant effect on sharpness. It does seem likely that some sharpness is lost from cine mode fitting so much dynamic range into so few bits. It may lose some of its ability to resolve low contrast detail as i believe someone has mentioned. still seems like something else is going on though... i dont think yet that the softness of cine mode is explained entirely in its lack of sharpening and low contrast.

Robert Ducon · October 10th, 2007, 12:25 PM

I did a near-uncompressed 4:2:2 capture-compare test a few months ago between TV and CINE.

I had the HV20 on sticks in a low-light environment, captured the feed uncompressed digitally and saved to the computer as ProRes HQ 422. Between the two PNG shots, I only changed the modes between TV and CINE - this makes it perfect for you to open the shots up in Photoshop, drop each onto a seperate layer, and switch between the layers looking for differences.

My own findings: a) more detail in TV mode b) less detail, and more latitude in CINE mode b) distracting grain in TV mode, next to no grain in CINE mode. The latitude in CINE mode is very apparent in these tests - to the left, you see that there is less blow out near the lamp's light, and you can ALSO see into the shadows under the television and speaker on the right in CINE.

EDIT: The uploads of these PNG files have failed 4x in a row.. I'm not over the size limit, so.. not sure what's up.

EDIT 2: I've spent 30 minutes trying to upload.. switched to smaller JPGs, didn't help.

EDIT 3: Finally.. the images are up.

Tip McPartland · October 10th, 2007, 01:24 PM

I didn't like my new HV20 much because of terrible grain in even moderate low light conditions. But when I tried Cinemode, the grain was so much improved that I now love the camera. I've been watching this thread to see what the tradeoff is under close evaluation, but to my reasonably experienced eye, if Cinemode gives greatly reduced noise and only slightly reduced resolution, I'm fine with that.

The question becomes: Is there a way to get the better resolution of TV mode and the noise reduction of Cinemode for the best of both worlds?

Tip McPartland

Michael Jouravlev · October 10th, 2007, 01:47 PM

Quote:

Originally Posted by Robert Ducon

My own findings: a) more detail in TV mode b) less detail, and more latitude in CINE mode b) distracting grain in TV mode, next to no grain in CINE mode. The latitude in CINE mode is very apparent in these tests - to the left, you see that there is less blow out near the lamp's light, and you can ALSO see into the shadows under the television and speaker on the right in CINE.

Judging by your pictures I did not notice more detail in Tv mode, it simply looks oversharpened. I bet that sharpening is a major contributor to the noisy picture.

My undestanding is that Cinemode is useless in 60i because the camera does not try to hold on to 1/60 shutter speed, it changes shutter speed to its liking. Cinemode plays nice only in 24p mode. I'll wait for HV30 in hope that they fix HV20's peculiarities.

Chris Barcellos · October 10th, 2007, 02:40 PM

Quote:

Originally Posted by Michael Jouravlev

Judging by your pictures I did not notice more detail in Tv mode, it simply looks oversharpened. I bet that sharpening is a major contributor to the noisy picture.

My undestanding is that Cinemode is useless in 60i because the camera does not try to hold on to 1/60 shutter speed, it changes shutter speed to its liking. Cinemode plays nice only in 24p mode. I'll wait for HV30 in hope that they fix HV20's peculiarities.

You snooze, you lose. If history is any lesson, my bet is that HV30 will have less professional capabilities than the HV20, more like tje HG10, and much like Sony did going from HC1 to HC3. They don't like us candidates for the A1 or H1 messing with their consumer cameras in place of those.

Michael Jouravlev · October 10th, 2007, 03:14 PM

Quote:

Originally Posted by Chris Barcellos

You snooze, you lose. If history is any lesson, my bet is that HV30 will have less professional capabilities than the HV20, more like tje HG10, and much like Sony did going from HC1 to HC3. They don't like us candidates for the A1 or H1 messing with their consumer cameras in place of those.

True, therefore my plan is actually to buy the HV20 sometime next January and keep it sitting in the box unopened. If something better comes up from Canon, I will just return the HV20 and will not pay restocking fee. If not, then I will keep the HV20. This will work for me because I don't really NEED a high-def camera right now, so I can wait. Also, there are so many HV20's around, I am pretty sure it we be available from eBay and other online retailers for some time.

In regards to HG10, what exactly is it missing? Manual audio level control? And a thumbwheel replaced with iPod-wheel? Does not seem to have lost a lot compared to the HV20.

The Cine screenshot above looks really smooth and nice. I wonder can Tv be brought closer by decreasing sharpening? Sharpening and contrast are related, cannot get sharp image with low contrast, so I hope that less sharpening can bring down noise as well as smoothen gamma curve.

---
Canon Elura User Pages

Fergus Anderson · October 10th, 2007, 03:40 PM

I tend to shoot cine and sharpen in post which suits most situations. I do use TV mode for specifying the shutter sometimes which is useful but even with -1 brightness and -1 contrast it looks blow out and more video like.

Chris Barcellos · October 10th, 2007, 04:13 PM

Quote:

Originally Posted by Michael Jouravlev

In regards to HG10, what exactly is it missing? Manual audio level control? And a thumbwheel replaced with iPod-wheel? Does not seem to have lost a lot compared to the HV20.

Canon Elura User Pages

While I haven't actually used the HG10, the issue is with focus. According to Chris Hurd, you have to have LCD open to focus, so if you are going viewfinder, that out. Even with LCD open, I wonder about exerting the pressure against the extended lcd, and how that will transmit to the axial movement of the camera. With the HV20, some of us added an adapter wheel to turn the thumb wheel to create an easier focus process.

Michael Jouravlev · October 10th, 2007, 04:14 PM

Quote:

Originally Posted by Fergus Anderson

I tend to shoot cine and sharpen in post which suits most situations. I do use TV mode for specifying the shutter sometimes which is useful but even with -1 brightness and -1 contrast it looks blow out and more video like.

If the HV20's controls are anything like on a standard TV, then to flatten gamma you actually need to increase brightness (black level), not to decrease it. By decreasing black level you simply are losing near-black information.

Noah Yuan-Vogel · October 10th, 2007, 08:26 PM

The HV20 controls are not like a standard tv where brightness is just black level. in the HV20, brightness controls how the camera sets exposure when autoexposure is on.

Euisung Lee · October 10th, 2007, 10:39 PM

Thanks Robert for doing the test. I agree with Michael and i don't think Tv mode actually holds more detail, but just has higher contrast. I suspect that if you look at the uncompressed screen the difference would be even smaller.

Dale Backus · October 10th, 2007, 11:53 PM

I know i'm kinda late coming in on this topic, but i just spent the last hour reading this thread from start to finish and i couldn't help saying a few things.

I was very surprised that this thread went as far as it did without trying it using uncompressed capture. IMO that is the only way to get what the sensor is actually trying to give us. So much more time was spent worrying about image compression - jpg over png - than the source compression! HDV is FAR worse than JPEG (obviously that depends on the level of JPEG compression).

HDV DESTROYS images.. it completely ruins them IMO. Talk about latitude, the first time i captured uncompressed and took it into post, i couldn't believe how much data was actually intact. It was mind-blowing. I don't know much about encoding algorithms and all that, but i do know that there is more than a substantial difference between 4:2:0 and 4:2:2. Keep in mind 1920 over 1440 and 100 mb/s over 3. Obviously the 100mb isn't necessary, but still.

I hooked up my HV20 and shot it at a faucet really close. I don't like grain or the HV20 trying to boost gain. So i set the camera to TV and set it to 1/48th, then pointed it at a light and locked the exposure. I then bumped the exposure up to open the aperture as wide as it would go. I wanted a clean noiseless image. I captured a couple seconds via HDMI fully uncompressed through my blackmagic intensity. Then i went and changed it to Cinemode. I dropped the exposure a bit to match the other test a little more and captured a couple more seconds.

I then captured full-res stills from the video. I put them into photoshop and checked and unchecked the layers over and over.

If anything... Cinemode has more detail - if there is any difference at all. The difference is so slight. You know the rest, you know what the differences are.. but know that no detail is "lost" either way. And if you're recording uncompressed or with a decent codec like cineform of prores or whatever at 4:2:2 8 bit, don't worry about losing information - that is unless of course you over or severely underexpose things.

But if you're recording to HDV... you have my prayers. (sorry to sound so melodramatic, but i've seen the light)

Here's a link for those who are interested... i may have jumbled things around, i may not have... but can you tell what is what? (Note: this isn't a chart or whatever, but who often do we film charts when we're making movies?)

www.dalebackus.com/allfoursmall.png

and if you really need the full res version (10mb)

www.dalebackus.com/allfour.png

Dale

Dale Backus · October 11th, 2007, 12:06 AM

An example of what i was talking about earlier...

The darker one is actually how it was captured. It was a fairly low-light situation and i had it on Tv mode 1/48th and exp +11 with no gain.

www.dalebackus.com/moescupuncorrected.png

Took it into photoshop and the only adjustment made was levels.

www.dalebackus.com/moescupuncorrected.png

Just fyi.

October 10th, 2007, 12:25 PM	#184
Robert Ducon Major Player Join Date: Nov 2006 Location: Victoria BC Posts: 400	HDMI Live 1080i Video Capture - TV vs Cine Modes I did a near-uncompressed 4:2:2 capture-compare test a few months ago between TV and CINE. I had the HV20 on sticks in a low-light environment, captured the feed uncompressed digitally and saved to the computer as ProRes HQ 422. Between the two PNG shots, I only changed the modes between TV and CINE - this makes it perfect for you to open the shots up in Photoshop, drop each onto a seperate layer, and switch between the layers looking for differences. My own findings: a) more detail in TV mode b) less detail, and more latitude in CINE mode b) distracting grain in TV mode, next to no grain in CINE mode. The latitude in CINE mode is very apparent in these tests - to the left, you see that there is less blow out near the lamp's light, and you can ALSO see into the shadows under the television and speaker on the right in CINE. EDIT: The uploads of these PNG files have failed 4x in a row.. I'm not over the size limit, so.. not sure what's up. EDIT 2: I've spent 30 minutes trying to upload.. switched to smaller JPGs, didn't help. EDIT 3: Finally.. the images are up. Attached Thumbnails __________________ Mac + Canon HV20 Last edited by Robert Ducon; October 10th, 2007 at 01:34 PM.

October 10th, 2007, 01:24 PM	#185
Tip McPartland Major Player Join Date: Nov 2005 Location: Crestline, California Posts: 351	Grain... I didn't like my new HV20 much because of terrible grain in even moderate low light conditions. But when I tried Cinemode, the grain was so much improved that I now love the camera. I've been watching this thread to see what the tradeoff is under close evaluation, but to my reasonably experienced eye, if Cinemode gives greatly reduced noise and only slightly reduced resolution, I'm fine with that. The question becomes: Is there a way to get the better resolution of TV mode and the noise reduction of Cinemode for the best of both worlds? Tip McPartland

October 1st, 2007, 04:39 PM	#183
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Actually the software used to do the test adds sharpening to the images to make them (ideally) equal in terms of digital sharpening. The result is the measured resolution gets much higher for cine mode and very slightly higher for the Tv stuff that is already sharpened more than cine. So it becomes more obvious that while cine is soft, it has almost as much actual resolution as Tv, but does still appear to be overall a bit less able to resolve detail. Once i get around to analyzing the latitude tests and gamma curves, it might be able to tell me a bit more about whether the contrast has a significant effect on sharpness. It does seem likely that some sharpness is lost from cine mode fitting so much dynamic range into so few bits. It may lose some of its ability to resolve low contrast detail as i believe someone has mentioned. still seems like something else is going on though... i dont think yet that the softness of cine mode is explained entirely in its lack of sharpening and low contrast.

October 10th, 2007, 03:40 PM	#189
Fergus Anderson Major Player Join Date: Nov 2006 Location: London, UK Posts: 321	I tend to shoot cine and sharpen in post which suits most situations. I do use TV mode for specifying the shutter sometimes which is useful but even with -1 brightness and -1 contrast it looks blow out and more video like.

October 10th, 2007, 08:26 PM	#192
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	The HV20 controls are not like a standard tv where brightness is just black level. in the HV20, brightness controls how the camera sets exposure when autoexposure is on.

October 10th, 2007, 10:39 PM	#193
Euisung Lee Regular Crew Join Date: Apr 2006 Location: San Francisco, CA Posts: 57	Thanks Robert for doing the test. I agree with Michael and i don't think Tv mode actually holds more detail, but just has higher contrast. I suspect that if you look at the uncompressed screen the difference would be even smaller.

October 10th, 2007, 11:53 PM	#194
Dale Backus Regular Crew Join Date: Nov 2006 Location: Cary, NC Posts: 132	I know i'm kinda late coming in on this topic, but i just spent the last hour reading this thread from start to finish and i couldn't help saying a few things. I was very surprised that this thread went as far as it did without trying it using uncompressed capture. IMO that is the only way to get what the sensor is actually trying to give us. So much more time was spent worrying about image compression - jpg over png - than the source compression! HDV is FAR worse than JPEG (obviously that depends on the level of JPEG compression). HDV DESTROYS images.. it completely ruins them IMO. Talk about latitude, the first time i captured uncompressed and took it into post, i couldn't believe how much data was actually intact. It was mind-blowing. I don't know much about encoding algorithms and all that, but i do know that there is more than a substantial difference between 4:2:0 and 4:2:2. Keep in mind 1920 over 1440 and 100 mb/s over 3. Obviously the 100mb isn't necessary, but still. I hooked up my HV20 and shot it at a faucet really close. I don't like grain or the HV20 trying to boost gain. So i set the camera to TV and set it to 1/48th, then pointed it at a light and locked the exposure. I then bumped the exposure up to open the aperture as wide as it would go. I wanted a clean noiseless image. I captured a couple seconds via HDMI fully uncompressed through my blackmagic intensity. Then i went and changed it to Cinemode. I dropped the exposure a bit to match the other test a little more and captured a couple more seconds. I then captured full-res stills from the video. I put them into photoshop and checked and unchecked the layers over and over. If anything... Cinemode has more detail - if there is any difference at all. The difference is so slight. You know the rest, you know what the differences are.. but know that no detail is "lost" either way. And if you're recording uncompressed or with a decent codec like cineform of prores or whatever at 4:2:2 8 bit, don't worry about losing information - that is unless of course you over or severely underexpose things. But if you're recording to HDV... you have my prayers. (sorry to sound so melodramatic, but i've seen the light) Here's a link for those who are interested... i may have jumbled things around, i may not have... but can you tell what is what? (Note: this isn't a chart or whatever, but who often do we film charts when we're making movies?) www.dalebackus.com/allfoursmall.png and if you really need the full res version (10mb) www.dalebackus.com/allfour.png Dale

October 11th, 2007, 12:06 AM	#195
Dale Backus Regular Crew Join Date: Nov 2006 Location: Cary, NC Posts: 132	An example of what i was talking about earlier... The darker one is actually how it was captured. It was a fairly low-light situation and i had it on Tv mode 1/48th and exp +11 with no gain. www.dalebackus.com/moescupuncorrected.png Took it into photoshop and the only adjustment made was levels. www.dalebackus.com/moescupuncorrected.png Just fyi.