Bad Image - Is This A Shutter Speed Or A Compression Problem?

This guy looks like he is in the transporter on Star Trek...


A couple of questions:

Is this a live view or recorded segment?

What is the shutter speed of the camera set to in this scene?

Or both? A couple other questions...

What is the camera model? Integrated IR?

The illumination on the left, is that natural light or IR?

Is the camera set at a fixed gain or fixed shutter? Priority assigned to which?

The window title is also compressed leading me to believe that the uploaded image has been additionally squashed, making it hard to evaluate the original smushing amount. Maybe upload one at orginal res?

I would look at the processing power of the PC viewing it, along with the robustness of the video display card. Other things to look at would be the speed and bandwidth capabilities of the network and switches, and the processing power of the recorder.

It's a Panasonic WV-NW502S; no integrated IR. This is recorded video in a still shot from this morning around 2am; I'm unable to find a shutter speed or gain setting, but compression is set at H.264.

The WV-NW502S is closely related to the WV-NP502 as shown by this incriminating photo:

Unfortunately, according to this episode of "Spot the Mistake", the NP502 is only well-suited for low-light situations when one is making deceptive marketing videos for a dominant illumination manufacturer.

John's discussion and subsequent remedies most likely apply to this camera as well. From looking at the image it appears relatively noise free (for a night shot), so one would assume that gain somehow can be increased, hopefully to the point where the shutter can be sped up enough to eliminate the ghosting.

Do you have a problem with this video during the day as well or only at night?

This camera was set at 1.3 MP a few weeks ago and I changed it to 3 MP. Could that have caused the problem?

Unfortunately, I don't have enough stored video to check recordings at 1.3 MP for comparison. I still haven't found a gain or shutter speed setting.

I have seen a problem like that similar when trying to use a specific camera driver in the VMS, but the password was mismatched on the VMS side and the camera side, so the VMS could not take full control of the camera. The resolution and FPS settings were also mismatched. When I corrected the password the problem went away, I theorized because then the VMS was able to sync settings with the camera with what it was expecting in the stream.

Should be in the image / privacy tab / pop-up, like so:

Also, what's the bitrate the camera is set to? It is going to be CBR so the camera might not have enough bandwidth for a night time, noisier scene.

This camera was set at 1.3 MP a few weeks ago and I changed it to 3 MP. Could that have caused the problem?

Since this is a new artifact, I would say if nothing else has changed then the odds on favorite would be the change from 1.3MP Super Dynamic to 3.0MP High-res mode is the culprit. The processing method of the 1.3MP is quite different than usual (relies on the fact the the sensor is a CCD type, not CMOS) as shown:

Although it is still an open question of whether the gain can be increased sufficiently to overcome the ghosting in the 3.0MP setting, thereby giving possibly the best overall result.

Well if the bitrate stayed the same (as these are CBR cameras), and you increased the resolution, quality / compression issues would more likely manifest.

True, though also the CBR can be setup to reduce the frame rate to maintain stasis and leave res and q untouched, no?

No.

Ok, I'm ready for my lesson.;)

as these are CBR cameras

Do you mean that they are only CBR? The spec indicates CBR is only one of 4 Transmission Priorities.

Well if the bitrate stayed the same

But why would it? The mode isnt flipping from 3.0 to 1.3 on its own (as far as I can tell) so it would likely transmit at one of the other, higher bit rates listed below, even if CBR was the only choice.

Transmission Priority: Constant bitrate / Frame rate priority / Best effort /Advanced VBR
Frame Rate: 1 / 3 / 5 / 7.5 /10 / 15 / 20* / 30* fps
* 1.3 Megapixel mode only
Bit Rate/Client: 64 / 128 / 256 / 384 / 512 / 768 / 1,024 / 1,536 / 2,048 / 3,072 / 4,096 / 8,192* kbps / Unlimited

In reality, they are all essentially CBR. The bit rate does not vary much at all, regardless of what you use. And to extent that you use frame priority mode, it will favor increasing compression more than decreasing frame rate but will fairly quickly start dropping frames as well if bandwidth is too constrained. It was weird to me when I first dealt with it, but a lot of these companies do not have 'true' VBR at all.

With Panasonic and with most CBR 'centric' IP cameras I have dealt with, bandwidth input is completely independent from resolution and frame rate. So if you drop resolution from 3MP to 4CIF and frame rate from 30 to 4fps, if the bitrate was originally set to 4Mb/s, it will simply (and dumbly) reduce compression and still stream at 4Mb/s, massively wasting bits. In this case, I am suggesting, it would be the opposite - the bit rate is now too low for the new increased resolution setting.

I'm assuming that your initial optimism at the delivery of the Panasonic VBR firmware update with 'best effort' proved to be unwarranted, and though the 'effort' may have been 'superlative' the result was anything but? Yes?

John Honovich Mod • 2 years ago

Thanks. I reviewed a Panasonic camera's web interface and could find bit rate and frame rate options. However, I could not find best effort. Where is that?...

Ascelikay • 2 years ago
Majority of i-Pro SmartHD cameras support this function. As the beginning of this Jan. NP502 also supports with firmware version update

John Honovich Mod • 2 years ago
Thanks. This looks to be a new and useful feature for Panasonic cameras. We upgraded a 502 and were able to confirm that with the most recent firmware it now has best effort. I've updated the post to reflect this news.

'Best effort' still operates very much like CBR.

Picture's a bit small, but from what I can see, it seems like a problem with dropped i-frames. Do you have a larger version?

There's blur that's KINDA like motion blur, but normally that would just be a solid smear, not a single clear image with "ghosts".

Edit: switching from 1.3MP to 3MP would certainly increase bitrate, and quite possibly lead to dropped frames.

...it seems like a problem with dropped i-frames...

If you turn out to be right, I will hereby nominate you for the first ever IPVM Sharpie award. ;)

Well, except I meant b-frame/p-frame... but same idea applies.

Thanks, Mr H! That's what I was looking for but the privacy tab wasn't bringing anything up. Now I see it!

I may need more help with the actual settings, but at least I can see them now.

I believe that camera only does JPEG in true 3 MP mode, so if the images above are 4:3 JPEG 3 MP, there would be no b/p/i frames. If in 1080 resolution 3 MP mode 16:9, it can do H.264. (This burned me on a project from a similar 3MP JPEG/1.3 MP H.264 Panasonic awhile back where I was proposing and sold a 3 MP camera, until I found out it had to be in JPEG mode). Not sure if the above image is a cropped 1920 x 1080 or 2048 x 1536. I never did figure this out and assumed that it had to do with their high resolution compression only supporting HD and UHD standards??

3 Megapixel mode (Aspect ratio [4:3]): H.264 :1,280 x 960 / VGA (640 x 480) / QVGA (320 x 240), up to 15 fps MPEG-4: VGA (640 x 480) / QVGA (320 x 240), up to 15 fps JPEG: 2,048 x 1536 / 1,280 x 960 / VGA (640 x 480), up to 15 fps 3 Megapixel mode (Aspect ratio [16:9]): H.264: 1,920 x 1,080 / 640 x 360 / 320 x 240, up to 15 fps JPEG: 1,920 x 1,080 / 640 x 360 / 320 x 240, up to 15 fps

I did notice that the artifacting or dematerializing (or whatever the term is) is happening on live view, so it's definitely not compression. Right?

Here are most of the settings on the camera if you are able to read them. I think they were probably the default settings for 1.3MP. Is there anything I can change to alleviate the issue?

Cynthia, that helps - a lot!

Shutter speed is 1/30s default so unless that guy was running, I doubt it's a shutter speed problem.

It could be a bit rate issue as the bit rate is set to 4Mb/s which might be too low at night. Try increasing the bit rate to the next level up.

Cynthia, the compression happens at the camera (not counting the VMS transcoding later), so it would affect live view equally.

Can you post a full-size version of the image?

Cynthia, did you change out of the 1.3MP mode to solve a particular problem? Or just on general principles of maximizing the camera's capabilities? Would it cause something else to break if you went back?

Also, based on the look of the GUI vs. the manual I'm using, it seems you might not have the latest firmware. @John, Is that something you think is worth updating if that's the case?

Well, all manufacturers recommend updating firmware :)

In this case, I think it shouldn't hurt.

Try changing DNR settings..... Really looks like DNR smearing.... //A

Thanks! Switching DNR to low is worth a shot.

Thanks so much for the suggestions!

I first changed the bit rate to the next level (8192 kbps) and the camera lost its connection to the server. When I changed it back, the camera reconnected. Then I changed the DNS from high to low. It seemed to help, but then a couple of runners came into view. I have attached the still photo.

I suppose a firmware update is in order, but I am a bit apprehensive as I have heard horror stories about updates gone awry. Does anyone have advice on firmware updates for cameras?

See, this looks like dropping P-frames to me. That increasing the bitrate causes loss of video altogether just reinforces that - it all suggests that you're having network communication issues. Whether that means the network is saturated, or there's some other issue, will require further troubleshooting... but maybe try dropping the bitrate further and see if the "transporter" effect goes away.

I don't know about these cameras, but some brands I've used change their bitrate when you change the framerate and/or resolution. Does dropping back to 1.3MP cause the bitrate setting to change as well?

See, this looks like dropping P-frames to me

So what do you want, a medal?

But actually it looks more like your first guess, dropped I-frame(s), unless he is running backwards ;)

Am I correct in assuming that if Cynthia does an export, we won't actually 'see' any missing frames because the export will generate bogus ones? Or maybe some sort of raw export would work to confirm?

Dropped I-frames tends to result in a completely grey or blank image with artifacts of the P-frames.

And no, you're not correct, because the frames would be lost between the camera and NVR, and so recorded exactly the way you see it.

And no, you're not correct, because the frames would be lost between the camera and NVR, and so recorded exactly the way you see it.

I didn't say anything about how it would be recorded, did I? What I was asking was the best way to investigate the recorded stream to see what frames were missing. I have seen DVR utilities that let you export in a variety of media-player friendly formats. My point was that if exported in this manner, the utility stream would actually re-encode the stream for the target selected and end up obscuring the original frame sequence. It would look the same as the recorded video of course, but would be a new transcoded sequence. Hence a raw export of the actual frames would be in order.

Dropped I-frames tends to result in a completely grey or blank image...

Just curious when would it blank out, when the first p-frame comes after a lost i-frame?

"Friends" some great attitude in here….. I will try to keep up the momentum...

Nice try, but no balloons……

This is clearly a result of a camera with “limited” lowlight capabilities using severe gain to give a usable picture and with all the artifacts a high temporal noise suppression/filtering will give you trying to remove the noise. This can be OK for verification/detection usage of the video but not for identification

This temporal/inter frame noise suppression will give more and more of these “ghosting’s” the worse the initial signal is. It is in the nature of the technology. If possible try to turn off the DNR completely and you will see none of this ghosting but instead massive noise (plague or cholera….)

I have “limited” experience with this specific camera model but a lot of trial and errors with MANY other ones…..

Exactly the same issues as in these pictures

These DNR settings (temporal) can be tuned against and together with spatial noise reduction (intra frame) that gives some “de focus” but less motion blur like the temporal ones. Sometimes this also referred to 2D noise reduction. If it is possible to turn them (temporal and spatial (inter frame, intra frame)) on and off and also change their individual settings you can often get a good result.

But if you are using a camera that don’t have the foundations for producing the picture from the beginning you are in a dead end.

Some manufacturer’s also do interesting stuff with the sensor and it’s pixels like binning to increase sensitivity. This can be a result when going from 1.3 to 3MP, increasing pixels=decreasing sensitivity=more gain=more noise=more temporal suppression=more ghosting.

Also some WDR “magic” can cause these kind of issues, especially when using different simultaneously shutter speeds in lowlight. Therefore many cameras turn off WDR in lowlight situations.

dual shutter speed WDR camera in lowlight with high temporal noise filtering can be scary......

Using a slower shutter can actually make this better since you increase sensitivity. Slow shutter blur is sometimes better (less destroying) than DNR blur, still not what I assume you prefer.

Suggestion:

  • Turn off DNR (if possible...)
  • Turn off WDR
  • Try 1/15 instead of 1/30 shutter speed to see if it looks better, not sure it will improve but worth a try if nothing else helps
  • Go back to 1.3 MP
  • Put up an IR illuminator

Still all compromises to fix the initial issue. Also correlating to earlier tests of resolution/picture quality, higher pixel count doesn’t always mean better result.

Btw, this have probably nothing to do with loosing frames (regardless of which one’s) or package disorder in the network, neither GPU rendering, that looks completely different from this noise reduction “inter frame ghosting” that we see here.

So, I hope I pointed this out as bluntly and ignorant as it was planned in a sarcastic way :) or?

//A

This is funny, I was just looking for an example from a test I did of spatial noise reduction. It's what it looks like to me, too.

Sorry, I'm all out of medals. ;)

Sorry all,

I missed the:

  • decreese AGC (gain)

It can also help a bit in the combined settings

//A

I tested this model camera a while back and was having the exact same issues. I tried every combination along with setting the camera to the highest available bitrate. I was also having this ussue during the day as well. I was working with the camera in 3MP mode and not 1.3MP because if I want'ed a 1.3MP camera, I would have purchased one for a lot less money.

I ended up calling Panasonic Tech Support and they walked through every setting with me and we never got any improvement. They ended up telling me that the camera really needs to run in 1.3MP mode and not at 3MP. Uhh, what's the point of paying top dollar for a 3MP camera to run it in 1.3MP mode? Just buy an awesome 1.3MP camera for a ton less.

Tech Support ended up telling me the camera must be defective and I needed to send it back. I sent it back alright, then scratched that model off the list of offerings to my clients.

Just my experience.

Seems to me that everyone is doing a lot of guessing, when the the simplest solution would be to add more light and see if that clears things up. Stick an IR on there, and if you don't get ghosting anymore, the answer was "shutter speed". If you still get ghosting, then the answer is "something else, possibly a compression problem, or maybe bandwidth, or something else".

Guessing is step one. Testing is step two. Repairing is step three. Billing is step four. Bragging about step three on IPVM is step five. Never leave any of the steps out.

Guessing is step one. Testing is step two...

Hold on Ari, this thread actually has a court-mandated twelve-step program and we're making progress! Recent contributions include:

Gloating (Courtesy of Matt)

See, this looks like dropping P-frames to me.

Taunting (Courtesy of Undisclosed A)

Nice try, but no balloons……

Bragging about Taunting (Also Courtesy of Undisclosed A)

So, I hope I pointed this out as bluntly and ignorant as it was planned in a sarcastic way :) or?

Reprimanding (Courtesy of Ari)

Never leave any of the steps out.

Mocking (Yours Truly)

Error: Out of memory - recursive inline quotation exception, exiting...